Final Report A Second Year Evaluation Study of Promethean ActivClassroom Prepared by Marzano Research Laboratory for Promethean, Ltd. December, 2010 MARZANO RESEARCH LABORATORY www.MarzanoResearch.com Phone: 888-849-0851 Fax: 866-308-3135 OUR MISSION To provide the best research, the most useful actions, and the highest level of services to educators. OUR VISION To continuously develop tools that translate high quality educational research into practical applications educators can put to immediate use. OUR GOAL To be the place educators go for the latest information and data, synthesized into clear, concise resources that facilitate immediate action. REPORT AUTHORS Robert J. Marzano & Mark W. Haystead For citation purposes, please refer to this document as: Marzano, R. J., & Haystead, M. W. (2010). Final report: A second year evaluation study of Promethean ActivClassroom. Englewood, CO: Marzano Research Laboratory. Copyright © 2010. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means without permission from Promethean, Ltd. http://www.prometheanworld.com. BUSINESS DEVELOPMENT OFFICE 555 N Morton Street Bloomington, IN 47404 Phone: 888-849-0851 Fax: 812-336-7790 RESEARCH CENTER 9000 E. Nichols Ave. Ste. 112 Englewood, CO 80112 Phone: 303-766-9199 Fax: 303-694-1778 www.MarzanoResearch.com Table of Contents Table of Contents ......................................................................................................................................... i Table of Figures ......................................................................................................................................... iii Executive Summary ................................................................................................................................... vi Introduction ..................................................................................................................................................1 Phase I ..........................................................................................................................................................2 Design of the Evaluation Study for Phase I ...........................................................................................3 The Use of Meta-Analysis .....................................................................................................................3 The Sample ............................................................................................................................................4 Demographics ........................................................................................................................................7 Data Analysis and Findings .................................................................................................................11 Evaluation Question 1: What effect does Promethean ActivClassroom have on students’ achievement regarding the subject matter content taught by their teachers?.................................12 Evaluation Question 2: Does the effect of Promethean ActivClassroom differ between school levels? ............................................................................................................................................20 Evaluation Question 3: Does the effect of Promethean ActivClassroom differ between grade levels? ............................................................................................................................................21 Evaluation Question 4: Does the effect of Promethean ActivClassroom differ between academic content areas? .................................................................................................................................23 Evaluation Question 5: Is there a relationship between the effect of Promethean ActivClassroom and length of teaching experience, how long teachers have used Promethean ActivClassroom, the percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ selfreported confidence in their use of Promethean ActivClassroom? ................................................24 Phase I Summary and Interpretation ....................................................................................................35 Phase II.......................................................................................................................................................38 Coding of Videotapes ..........................................................................................................................40 Evaluation Question 1: What is the relationship between student engagement and positive versus negative effects for the use of Promethean ActivClassroom? .......................................................41 Evaluation Question 2: What is the relationship between teacher IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? .......................................................43 Evaluation Question 3: What is the relationship between student IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? .......................................................44 Evaluation Question 4: What is the relationship between multiple student use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? ..............................46 Evaluation Question 5: What is the relationship between student independent use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? ..............................47 Final Report i 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 6: What is the relationship between teacher use of IWB reinforcers and positive versus negative effects for the use of Promethean ActivClassroom? ..............................48 Evaluation Question 7: What is the relationship between teacher use of learner response systems and positive versus negative effects for the use of Promethean ActivClassroom?........................49 Evaluation Question 8: What is the relationship between teacher use of the IWB to represent knowledge graphically or nonlinguistically and positive versus negative effects for the use of Promethean ActivClassroom? ........................................................................................................51 Evaluation Question 9: What is the relationship between teacher previewing of content and positive versus negative effects for the use of Promethean ActivClassroom? ..............................52 Evaluation Question 10: What is the relationship between teacher chunking of content and positive versus negative effects for the use of Promethean ActivClassroom? ..............................53 Evaluation Question 11: What is the relationship between teacher scaffolding of content and positive versus negative effects for the use of Promethean ActivClassroom? ..............................55 Evaluation Question 12: What is the relationship between teacher pacing of content and positive versus negative effects for the use of Promethean ActivClassroom? ............................................56 Evaluation Question 13: What is the relationship between teacher monitoring of student progress and positive versus negative effects for the use of Promethean ActivClassroom?........................57 Evaluation Question 14: What is the relationship between clarity of content as depicted in the IWB and positive versus negative effects for the use of Promethean ActivClassroom? ...............59 Evaluation Question 15: What is the relationship between student interaction about the content and positive versus negative effects for the use of Promethean ActivClassroom?........................60 Evaluation Question 16: What is the relationship between student response rates and positive versus negative effects for the use of Promethean ActivClassroom? ............................................61 Evaluation Question 17: What is the relationship between classroom management and positive versus negative effects for the use of Promethean ActivClassroom? ............................................63 Interpretation ........................................................................................................................................64 Comparison of Means versus Bivariate Correlation ......................................................................64 Engagement and Student Achievement .........................................................................................64 Variables 6, 7, and 8 ......................................................................................................................65 Variables 10, 11, 12, 13, 14, and 16 ..............................................................................................65 Phase II Summary and Interpretation ..................................................................................................68 Technical Notes .........................................................................................................................................71 Appendix A – MRL Action Research Instructions ....................................................................................79 Appendix B – NCES Code Definitions .....................................................................................................82 Appendix C – Treatment/Control Study Data ...........................................................................................83 References ..................................................................................................................................................90 Final Report ii 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Table of Figures Figure 1. Participating Sites – Midwest Region ..........................................................................................5 Figure 2. Participating Sites – Northeast Region .........................................................................................5 Figure 3. Participating Sites – Southeast Region .........................................................................................5 Figure 4. Participating Sites – Southwest Region........................................................................................6 Figure 5. Participating Sites – West Region ................................................................................................6 Figure 6. Number of Participating Sites by School Level ...........................................................................6 Figure 7. Number of Students by Grade Level ............................................................................................7 Figure 8. Population Statistics and Locale Codes for Participating Sites ....................................................8 Figure 9. Number of Students by Study Locale Category – City ................................................................9 Figure 10. Number of Students by Study Locale Category – Suburb .........................................................9 Figure 11. Number of Students by Study Locale Category – Town ............................................................9 Figure 12. Number of Students by Study Locale Category – Rural ..........................................................10 Figure 13. Number of Students by Demographic Category – Gender .......................................................10 Figure 14. Number of Students by Demographic Category – Ethnicity ....................................................10 Figure 15. Number of Students by Demographic Category – Free & Reduced Lunch .............................11 Figure 16. Number of Students by Demographic Category – English Language Learner ........................11 Figure 17. Number of Students by Demographic Category – Special Education .....................................11 Figure 18. ANCOVA Findings for Treatment/Control Studies .................................................................13 Figure 19. Overall Random Effects ...........................................................................................................16 Figure 20. Homogeneity Analysis .............................................................................................................17 Figure 21. Distribution of Effect Sizes – Continuation Study ...................................................................18 Figure 22. Distribution of Effect Sizes – Combined Continuation and First Year Studies .......................19 Figure 23. Random Effects – School Level ...............................................................................................20 Figure 24. Homogeneity Analysis – School Level ....................................................................................20 Figure 25. Random Effects – Grade Level ................................................................................................21 Figure 26. Homogeneity Analysis – Grade Level .....................................................................................22 Figure 27. Random Effects – Academic Content Area..............................................................................23 Figure 28. Homogeneity Analysis – Academic Content Area ...................................................................24 Figure 29. Distribution of Teaching Experience – Continuation Study ....................................................25 Figure 30. Distribution of Teaching Experience – Combined Continuation and First Year Studies ........26 Figure 31. Distribution of Promethean ActivClassroom Use – Continuation Study .................................27 Final Report iii 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 32. Distribution of Promethean ActivClassroom Use – Combined Continuation and First Year Studies ........................................................................................................................................................28 Figure 33. Distribution of Percentage of Class Time – Continuation Study .............................................29 Figure 34. Distribution of Percentage of Class Time – Combined Continuation and First Year Studies .30 Figure 35. Distribution of Confidence Ratings – Continuation Study ......................................................31 Figure 36. Distribution of Confidence Ratings – Combined Continuation and First Year Studies ..........32 Figure 37. Product Moment Correlations for Teacher Survey Responses – Continuation Study .............33 Figure 38. Product Moment Correlations for Teacher Survey Responses – Combined Continuation and First Year Studies .......................................................................................................................................34 Figure 39. Percentage of Second Ratings the Same as First Rating and One Score Point from First Rating .........................................................................................................................................................40 Figure 40. Comparison of Means for Student Engagement .......................................................................41 Figure 41. Bivariate Correlation for Student Engagement ........................................................................42 Figure 42. Comparison of Means for Teacher IWB Skills ........................................................................43 Figure 43. Bivariate Correlation for Teacher IWB Skills ..........................................................................44 Figure 44. Comparison of Means for Student IWB Skills .........................................................................44 Figure 45. Bivariate Correlation for Student IWB Skills ..........................................................................45 Figure 46. Comparison of Means for Multiple Student Use of the IWB ...................................................46 Figure 47. Bivariate Correlation for Multiple Student Use of the IWB ....................................................47 Figure 48. Comparison of Means for Student Independent Use of the IWB .............................................47 Figure 49. Bivariate Correlation for Student Independent Use of the IWB ..............................................48 Figure 50. Comparison of Means for Use of IWB Reinforcers .................................................................48 Figure 51. Bivariate Correlation for Teacher Use of IWB Reinforcers .....................................................49 Figure 52. Comparison of Means for Use of Learner Response Systems .................................................50 Figure 53. Bivariate Correlation for Use of Learner Response Systems ...................................................50 Figure 54. Comparison of Means for Use of the IWB for Nonlinguistic Representation .........................51 Figure 55. Bivariate Correlation for Use of the IWB for Nonlinguistic Representation ...........................52 Figure 56. Comparison of Means for Previewing of Content ....................................................................52 Figure 57. Bivariate Correlation for Previewing of Content .....................................................................53 Figure 58. Comparison of Means for Teacher Chunking ..........................................................................54 Figure 59. Bivariate Correlation for Teacher Chunking ............................................................................54 Figure 60. Comparison of Means for Teacher Scaffolding .......................................................................55 Figure 61. Bivariate Correlation for Teacher Scaffolding .........................................................................56 Figure 62. Comparison of Means for Pacing of Content ...........................................................................56 Figure 63. Bivariate Correlation for Pacing of Content .............................................................................57 Final Report iv 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 64. Comparison of Means for Monitoring Student Progress ..........................................................58 Figure 65. Bivariate Correlation for Monitoring Student Progress ...........................................................58 Figure 66. Comparison of Means for Clarity of Content on the IWB .......................................................59 Figure 67. Bivariate Correlation for Clarity of Content on the IWB .........................................................60 Figure 68. Comparison of Means for Student Interaction .........................................................................60 Figure 69. Bivariate Correlation for Student Interaction ...........................................................................61 Figure 70. Comparison of Means for Student Response Rate ...................................................................62 Figure 71. Bivariate Correlation for Student Response Rate .....................................................................62 Figure 72. Comparison of Means for Classroom Management .................................................................63 Figure 73. Bivariate Correlation for Classroom Management...................................................................64 Figure 74. Product Moment Correlations between Variables 10, 11, 12, 13, 14, and 16 ..........................67 Figure 75. Weighted Meta-Regression Analysis with Corrected Effect Size as the Dependent Variable and Variables 10, 11, 12, 13, 14, and 16 as Independent Variables ..........................................................68 Final Report v 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Executive Summary During the 2009/2010 school year, Marzano Research Laboratory (MRL) was commissioned by Promethean Ltd. to conduct a second year evaluation study of the effects of Promethean ActivClassroom on student academic achievement. The findings from the first year evaluation study conducted during the 2008/2009 school year were reported in Final Report: An Evaluation Study of the Effects of Promethean ActivClassroom on Student Achievement (Marzano & Haystead, 2009). The weighted average effect size estimate (Cohen’s d) for 85 treatment/control studies was .37 and statistically significant at the .0001 level (p < .0001). When corrected for attenuation, the weighted average effect size (Cohen’s d) was .44 and statistically significant at the .0001 level (p < .0001). The second year evaluation study was conducted in two phases. Phase I involved an analysis of student learning with and without Promethean ActivClassroom as it relates to demographic information and teacher self-reported experience with the technology. Phase II involved an analysis of student learning with and without Promethean ActivClassroom as it relates to teacher behaviors as evidenced in videotapes of teachers using the technology in their classrooms. In the treatment group, teachers used Promethean ActivClassroom to augment their instructional practices. In the control group, teachers used strategies and materials to facilitate instruction without the use of Promethean ActivClassroom. Because students could not be randomly assigned to treatment and control groups, the second year evaluation study employed a quasi-experimental design, referred to as a pretest-posttest non-equivalent groups design. The pretest scores were used as a covariate to statistically equate the students and partially control for differing levels of background knowledge and skill. Phase I of the second year evaluation study attempted to answer the following questions through a metaanalysis of the treatment/control studies: Evaluation Question 1: What effect does Promethean ActivClassroom have on students’ achievement regarding the subject matter content taught by their teachers? Evaluation Question 2: Does the effect of Promethean ActivClassroom differ between school levels? Evaluation Question 3: Does the effect of Promethean ActivClassroom differ between grade levels? Evaluation Question 4: Does the effect of Promethean ActivClassroom differ between academic content areas? Evaluation Question 5: Is there a relationship between the effect of Promethean ActivClassroom and length of teaching experience, how long teachers have used Promethean ActivClassroom, the percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ self-reported confidence in their use of Promethean ActivClassroom? The weighted average effect size (Cohen’s d) for 46 treatment/control studies was .34 and statistically significant at the .01 level (p < .01). When the 46 effect sizes (Cohen’s d) were corrected for attenuation and combined, the weighted average effect size (Cohen’s d) was .39 and statistically significant at the .01 level (p < .01). No significant moderator effects (p < .05) were found for the following variables: school level, grade level, and academic content area. No significant correlations (p < .05) were found between the 46 effect sizes (Cohen’s d) for the treatment/control condition (i.e., use of Promethean ActivClassroom) and the following variables: length of teaching experience (reported in years), how Final Report vi 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 long teachers have used Promethean ActivClassroom (reported in months), the percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ self-reported confidence in their use of Promethean ActivClassroom. However, a significant correlation (p < .001) of .57 was found between teachers’ self-reported confidence in their use of Promethean ActivClassroom and how long teachers have used Promethean ActivClassroom (reported in months). When the results from the 46 treatment/control studies were combined with 85 treatment/control studies from the first year evaluation study of ActivClassroom, the weighted average effect size (Cohen’s d) was .36 and significant at the .0001 level (p < .0001). When the 131 effect sizes (Cohen’s d) were corrected for attenuation and combined, the weighted average effect size (Cohen’s d) was .41 and significant at the .0001 level (p < .0001). The combined data indicated no significant moderator effects (p < .05) for the following variables: school level, grade level, and academic content area. No significant correlations (p < .05) were found between the 131 effect sizes (Cohen’s d) for the treatment/control condition (i.e., use of Promethean ActivClassroom) and the following variables: length of teaching experience (reported in years), how long teachers have used Promethean ActivClassroom (reported in months), and teachers’ self-reported confidence in their use of Promethean ActivClassroom. However, a significant correlation (p < .05) of .22 was found between the effect sizes and the percentage of time Promethean ActivClassroom was used in the classroom. In addition, a significant correlation (p < .001) of .50 was found between teachers’ self-reported confidence in their use of Promethean ActivClassroom and how long Promethean ActivClassroom was used (reported in months), a significant correlation (p < .001) of .45 was found between teachers’ self-reported confidence in their use of Promethean ActivClassroom and the percentage of time Promethean ActivClassroom was used in the classroom, and a significant correlation (p < .05) of -.21 was found between teachers’ self-reported confidence in their use of Promethean ActivClassroom and the length of teaching experience (reported in years). It is useful to note that the combined results across the two evaluation studies involved 4,913 students, 123 teachers, 73 schools, and 36 districts. Phase II of the evaluation study sought to answer the following questions: Evaluation Question 1: What is the relationship between student engagement and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 2: What is the relationship between teacher IWB skills and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 3: What is the relationship between student IWB skills and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 4: What is the relationship between multiple student use of the IWB and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 5: What is the relationship between student independent use of the IWB and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 6: What is the relationship between teacher use of IWB reinforcers and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 7: What is the relationship between teacher use of learner response systems and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 8: What is the relationship between teacher use of the IWB to represent knowledge graphically or nonlinguistically and positive versus negative effects for use of Promethean ActivClassroom? Final Report vii 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 9: What is the relationship between teacher previewing of content and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 10: What is the relationship between teacher chunking of content and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 11: What is the relationship between teacher scaffolding of content and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 12: What is the relationship between teacher pacing of content and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 13: What is the relationship between teacher monitoring of student progress and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 14: What is the relationship between clarity of content as depicted in the IWB and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 15: What is the relationship between student interaction about the content and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 16: What is the relationship between student response rates and positive versus negative effects for use of Promethean ActivClassroom? Evaluation Question 17: What is the relationship between classroom management and positive versus negative effects for use of Promethean ActivClassroom? Again, Phase II involved an analysis of student learning with and without Promethean ActivClassroom from the perspective of teacher behavior as evidenced in videotapes of teachers using the technology in their classrooms. The general focus of Phase II was to determine the behaviors that differentiated those teachers who obtained positive effects from Promethean ActivClassroom from those who did not. Two types of analyses were performed for each of the 17 questions. One analysis separated the teachers into two categories: those with positive effects and those with negative effects for use of Promethean ActivClassroom. Membership in these groups was considered the independent variable and scores on the variables for the 17 Phase II evaluation questions were considered the dependent variables. An independent samples t-test was conducted comparing the mean for the group with positive effects and the mean for the group with negative effects for each of the 17 Phase II variables. The second analysis correlated scores on the 17 Phase II variables with the corrected effect sizes from Phase I. When the results from the first and second year evaluation studies were combined, the correlations between 15 of the 17 variables addressed in Phase II and the corrected effect sizes calculated in Phase I were statistically significant at the .05 level (p < .05). The correlations would also be considered significant at the .01 level (p < .01) for all 15 variables and at the .001 level (p < .001) for 12 of the 15 variables. Student IWB skills (Variable 3) and student independent use of the IWB (Variable 5) did not have significant correlations. Of the 15 Phase II variables with significant correlations to effect size, only classroom management (Variable 17) did not exhibit a statistically significant difference on the ttest. However, it should be noted that the t-test approached significance at the .05 level (t = 1.94, df = 98, p = .056). The correlations between 14 of the 17 Phase II variables and the corrected effect sizes were greater than .30. Of those 14 variables, chunking (Variable 10), scaffolding (Variable 11), pacing (Variable 12), monitoring (Variable 13), clarity of IWB (Variable 14), and student response rate (Variable 16) exhibited correlations with corrected effect size that were greater than .60. A multiple correlation of .789 (p < .0001) was found between 99 corrected effect sizes and these 6 variables. Final Report viii 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Introduction During the 2009/2010 school year, Marzano Research Laboratory (MRL) was commissioned by Promethean Ltd. to conduct a second year evaluation study of the effects of Promethean ActivClassroom on student academic achievement (hereinafter referred to as the continuation study). The findings from the first year evaluation study were reported in Final Report: An Evaluation Study of the Effects of Promethean ActivClassroom on Student Achievement (Marzano & Haystead, 2009). The weighted average effect size estimate (Cohen’s d) was statistically significant (d = .37, N = 85, p < .0001). When corrected for attenuation, the percentile gain associated with the use of Promethean ActivClassroom was 17 (d = .44, N = 85, p < .0001). None of the teachers, schools, or districts received compensation from Promethean or MRL, their participation was strictly voluntary. In March 2009 an invitation was sent to 28 school districts in various parts of the country soliciting volunteer teachers willing to participate in a second round of treatment/control studies as part of the first year evaluation study. Due to the timing of the data collection, 40 treatment/control studies were not included in the first year study. In November 2009 an invitation was sent to 40 school districts throughout the country (including some districts in the March 2009 invitation) seeking volunteer teachers for participation in the continuation study and an additional evaluation study focusing on Promethean Learner Response Systems. The 40 second round treatment/control studies from the March 2009 invitation were combined with 6 treatment/control studies from the November 2009 invitation as part of the continuation study. The specifics of the Promethean technologies available for use in the classroom can be found at Promethean’s website (see http://www.prometheanworld.com). Final Report 1 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Phase I Final Report 2 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Design of the Evaluation Study for Phase I At a basic level, the continuation study sought to determine the effect that Promethean ActivClassroom had on students’ achievement regarding academic content taught by their teachers. In more specific terms, the continuation study employed one primary independent variable, whether teachers used Promethean ActivClassroom to augment their current instructional practices (referred to as the treatment group) or did not use Promethean ActivClassroom to augment their current instructional practices (referred to as the control group). The dependent variable was students’ knowledge of academic content addressed during a unit of instruction. Secondary independent variables of interest—students’ school level, students’ grade level, academic content area—were treated as moderators. Meta-analytic techniques (explained below) were employed to determine if the effect differed between groups on those moderators. One aspect of this evaluation study worth mentioning is that each treatment/control study used a quasiexperimental research design. Teachers acted as their own control and collected pretest and posttest scores for two groups of students—one treatment and one control. The pretest scores were used as a covariate to statistically equate the students and partially control for differing levels of background knowledge and skill. Analysis of covariance (ANCOVA) is often used when random assignment is not feasible (see Technical Note 1). This approach was used to mitigate potential differences in teaching experience, style, methodology, etc. as possible explanations for differences between treatment and control groups. Unlike experimental designs, quasi-experimental designs lack random assignment. Quasi-experimental designs are often used when it is not possible or practical to randomly assign subjects to equivalent groups. Although ANCOVA was used to statistically equate students in terms of prior achievement, without randomization arguments about causal relationships are severely weakened. (For a more thorough discussion of the treatment/control study design and its implications, see Technical Note 2.) Teachers were instructed to teach a short unit (or set of related lessons) on a topic of their choice to two groups of students—one treatment and one control. Detailed directions provided to teachers can be found in Appendix A. Briefly though, instructional activities in both groups were to be as similar as possible except for the fact that Promethean ActivClassroom was used in one group only (i.e., the treatment group). At the elementary grades—particularly the lower elementary grades—if two intact groups of students could not be identified teachers were instructed to teach two separate units to the same group of students ensuring that the subject matter of each unit was as closely related as possible (i.e., complementary topics). The Use of Meta-Analysis Meta-analytic techniques (see Hedges & Olkin, 1985; Lipsey & Wilson, 2001; Cooper, 2009) were employed to aggregate the findings from the treatment/control studies using the statistical software package Comprehensive Meta-Analysis (CMA), v2.2.050 (Borenstein, Hedges, Higgins, & Rothstein, 2005). Meta-analytic techniques are often used to combine the results of studies on a common topic. Final Report 3 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 In the continuation study, a common class of interventions was used in all treatment classes (i.e., use of Promethean ActivClassroom). However, the studies employed teacher-designed pre- and postassessments of student academic achievement across various grade levels and academic content areas requiring different dependent measures. A cautionary note must be made here. As mentioned previously, all the treatment/control studies involved in the continuation study employed intact groups (i.e., students were not randomly assigned to treatment/control conditions). Although ANCOVA was used to statistically equate students in terms of prior academic achievement based on teacher-designed preassessments, arguments about causal relationships are not nearly as strong as they might be if group membership had been determined through random assignment (see Technical Note 2). To combine studies that used different dependent measures, the results of each study are translated into an effect size. While there are several different effect sizes, the one used in this report is the standardized mean difference effect size (Cohen’s d). A standardized mean difference is the difference between the treatment and control group mean scores expressed in standard deviation units. By convention, a positive effect size favors the treatment group. (For a discussion of effect size, see Technical Note 3.) Meta-analytic findings are typically reported in two ways: 1. Weighted mean effect sizes computed from estimates of effect size for each treatment/control study. 2. Weighted mean effect sizes computed from estimates of effect size for each treatment/control study that have been corrected for attenuation due to lack of reliability in the dependent measure (i.e., teacher-designed assessments of student academic achievement). Technical Note 4 explains the method used to correct for attenuation and an interpretation of such corrections. It is commonly recommended that estimates of effect size should be corrected for attenuation due to unreliability of the dependent measure (for a detailed discussion of attenuation, see Hunter & Schmidt, 2004). In basic terms, every assessment is imprecise to some extent. This imprecision lowers the estimate of the true effect size. Throughout this report, observed and corrected effect sizes are typically displayed for comparison. When this is the case, the discussion of findings is limited to the corrected results only. Although meta-analytic techniques were used to analyze most of the data, when more appropriate, the general linear model was employed using the statistical software package, PASW® Statistics (SPSS), v17.0.2 (SPSS, 2009). Specifically, when independent variables were more continuous in nature than categorical, the general linear model was commonly employed. The Sample As Figures 1 through 5 illustrate, 27 public schools participated in the second year evaluation study from 20 cities in 14 states. The sites are grouped by their location in one of five geographic regions— Midwest, Northeast, Southeast, Southwest, and West. Figures 1 through 5 also show the number of teachers involved in the treatment/control studies at each site. Final Report 4 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 1. Participating Sites – Midwest Region Site City State # of Teachers 1 Orchard Lake MI 1 2 St. Louis MO 4 3 Evansville IN 1 4 Evansville IN 1 5 Evansville IN 2 6 Evansville IN 1 10 Total Figure 2. Participating Sites – Northeast Region Site City State # of Teachers 7 Groton CT 1 8 New York NY 1 9 Geneva NY 5 Total 7 Figure 3. Participating Sites – Southeast Region Site City State # of Teachers 10 Sarasota FL 1 11 Sulphur LA 7 12 Lake Charles LA 3 13 Suwanee GA 1 14 Cumming GA 1 15 Cumming GA 1 16 Vestavia Hills AL 2 17 Vestavia Hills AL 2 18 Orlando FL 1 Total 19 Final Report 5 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 4. Participating Sites – Southwest Region Site City State # of Teachers 19 Round Rock TX 1 20 Pflugerville TX 1 21 Pflugerville TX 1 Total 3 Figure 5. Participating Sites – West Region Site City State # of Teachers 22 Snoqualmie WA 1 23 Madera CA 1 24 Wasilla AK 1 25 Palmer AK 1 26 Palmer AK 1 27 Willow AK 1 Total 6 Figures 1 through 5 indicate the continuation study involved 27 sites. In the Midwest region, 10 teachers participated at 6 sites; in the Northeast region, 7 teachers participated at 3 sites; in the Southeast region, 19 teachers participated at 9 sites; in the Southwest region, 3 teachers participated at 3 sites; and in the West region, 6 teachers participated at 6 sites. Figure 6 displays the number of participating sites by school level along with the number of students in control and treatment groups. Figure 6. Number of Participating Sites by School Level School Level # of Sites Control N Treatment N Total N Elementary School (Grades K – 5) 18 497 530 1,027 Middle School (Grades 6 – 8) 4 186 199 385 High School Grades (9 – 12) 5 86 77 163 Total 27 769 806 1,575 Final Report 6 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 The first year evaluation study involved 3,338 students—1,622 students in the control group and 1,716 students in the treatment group (Marzano & Haystead, 2009). The continuation study involved 1,575 students—769 students in the control group and 806 students in the treatment group. Of those students in the continuation study, 1,027 were at 18 sites that teach students at the elementary school level, 385 were at 4 sites that teach students at the middle school level, and 163 were at 5 sites that teach students at the high school level. For comparison purposes, teachers were asked to report the grade level(s) taught in their treatment/control studies. Figure 7 depicts the number of control and treatment students for each grade level in the second year continuation study. Two teachers did not report a grade level. In addition, three of the high school teachers reported more than one grade. Therefore, student counts for those teachers were excluded from Figure 7. Figure 7. Number of Students by Grade Level Grade Level Control N Treatment N Total N K 23 33 56 1 116 119 235 2 57 71 128 3 73 81 154 4 126 128 254 5 84 83 167 6 20 19 39 7 68 81 149 8 98 99 197 9 25 18 43 10 -- -- -- 11 -- -- -- 12 -- -- -- Total 690 732 1,422 The previous two figures (Figure 6 and Figure 7) indicate that 153 students were excluded from Figure 7—79 from the control group and 74 from the treatment group. Only student counts from treatment/control studies which identified a single grade level were included in Figure 7. Demographics To determine what type of community in which each participating school is located, MRL used data from the National Center for Education Statistics (NCES), a federal entity located within the U.S. Department of Education’s Institute of Education Sciences (see http://www.nces.ed.gov/). MRL utilized the NCES website to search its online databases for each of the 27 participating schools involved in the second year continuation study. Specifically, a search was conducted using the NCES Common Core of Final Report 7 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Data (CCD) “Search for Public Schools” webpage (http://nces.ed.gov/ccd/schoolsearch/). CCD public school data is currently available from the NCES website for the 2008-2009 school year. MRL also used the U.S. Census Bureau’s American FactFinder website to search for population statistics from the 2000 U.S. Census (http://factfinder.census.gov/). Figure 8 displays the city population statistics for each site, along with the NCES locale code and category assigned to that site. The locale codes are assigned based on the proximity of a site’s address to an urbanized area (see http://nces.ed.gov/ccd/rural_locales.asp). A listing of NCES site code definitions can be found in Appendix B. Figure 8. Population Statistics and Locale Codes for Participating Sites Site Locale Code Locale Category City State Population (2000 Census) 1 21 Suburb Orchard Lake MI 2,215 2 21 Suburb St. Louis MO 348,189 3 12 City Evansville IN 121,582 4 12 City Evansville IN 121,582 5 12 City Evansville IN 121,582 6 12 City Evansville IN 121,582 7 22 Suburb Groton CT 39,907 8 11 City New York NY 8,008,278 9 32 Town Geneva NY 13,617 10 13 City Sarasota FL 52,715 11 22 Suburb Sulphur LA 20,512 12 13 City Lake Charles LA 71,757 13 21 Suburb Suwanee GA 8,725 14 21 Suburb Cumming GA 4,220 15 41 Rural Cumming GA 4,220 16 41 Rural Vestavia Hills AL 24,476 17 41 Rural Vestavia Hills AL 24,476 18 12 City Orlando FL 185,951 19 21 Suburb Round Rock TX 61,136 20 21 Suburb Pflugerville TX 16,335 21 21 Suburb Pflugerville TX 16,335 22 31 Town Snoqualmie WA 1,631 23 13 City Madera CA 43,207 Final Report 8 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Site Locale Code Locale Category City State Population (2000 Census) 24 32 Town Wasilla AK 5,469 25 32 Town Palmer AK 4,533 26 32 Town Palmer AK 4,533 27 43 Rural Willow AK 1,658 Sources: From NCES Common Core of Data (http://nces.ed.gov/ccd/schoolsearch/) and U.S. 2000 Census (http://factfinder.census.gov/) . Figures 9 through 12 list the number of control and treatment students in the second year continuation study for each of the 12 NCES urban-centric locale codes. Figure 9. Number of Students by Study Locale Category – City Locale Code & Category Control N Treatment N Total N # of Sites 11 – City, Large 23 23 46 1 12 – City, Midsize 149 133 282 5 13 – City, Small 55 66 121 3 Total 227 222 449 9 Figure 10. Number of Students by Study Locale Category – Suburb Locale Code & Category Control N Treatment N Total N # of Sites 21 – Suburb, Large 170 178 348 7 22 – Suburb, Midsize 135 149 284 2 23 – Suburb, Small -- -- -- -- Total 305 327 632 9 Figure 11. Number of Students by Study Locale Category – Town Locale Code & Category Control N Treatment N Total N # of Sites 31 – Town, Fringe 29 23 52 1 32 – Town, Distant 102 124 226 4 33 – Town, Remote -- -- -- -- Total 131 147 278 5 Final Report 9 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 12. Number of Students by Study Locale Category – Rural Locale Code & Category Control N Treatment N Total N # of Sites 41 – Rural, Fringe 94 95 189 3 42 – Rural, Distant -- -- -- -- 43 – Rural, Remote 12 15 27 1 Total 106 110 216 4 As Figures 9 through 12 illustrate, there are 449 students from 9 sites located in cities, 632 students from 9 sites located in suburban areas, 278 students from 5 sites located in towns, and 216 students from 4 sites located in rural areas. Figures 13 through 17 list the number of control and treatment students in the second year continuation study for each demographic category. Figure 13. Number of Students by Demographic Category – Gender Control N Treatment N Total N Male 343 347 690 Female 325 345 670 Not Reported 101 114 215 Total 769 806 1,575 Figure 14. Number of Students by Demographic Category – Ethnicity Control N Treatment N Total N Asian 19 34 53 African American 81 85 166 Hispanic 54 55 109 Native American 3 5 8 Other 23 23 46 White/Caucasian 503 531 1,034 Not Reported 86 73 159 Total 769 806 1,575 Final Report 10 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 15. Number of Students by Demographic Category – Free & Reduced Lunch Control N Treatment N Total N Yes 220 219 439 No 409 431 840 Not Reported 140 156 296 Total 769 806 1,575 Figure 16. Number of Students by Demographic Category – English Language Learner Control N Treatment N Total N Yes 91 97 188 No 585 602 1,187 Not Reported 93 107 200 Total 769 806 1,575 Figure 17. Number of Students by Demographic Category – Special Education Control N Treatment N Total N Yes 70 58 128 No 604 636 1,240 Not Reported 95 112 207 Total 769 806 1,575 Data Analysis and Findings One dependent variable was considered in the second year continuation study: students’ knowledge of academic content addressed by their teacher during a unit of instruction. The primary independent variable of interest was the treatment/control condition—whether students were exposed to Promethean ActivClassroom or not. Also of interest were the differences in potential effect of Promethean ActivClassroom with respect to three moderator variables—school level, grade level, academic content area. The control condition represented the aggregate strategies and materials used by the teachers to facilitate instruction without the use of Promethean ActivClassroom. Surveys submitted by participating teachers indicated that they were quite diverse in the approaches they used. As mentioned previously, two statistical software packages were utilized for data analysis. Data from each treatment/control study was first analyzed with the general linear model using SPSS. One independent variable (treatment/control condition) was entered into the equation using a fixed-effects model (see Technical Note 5 for a discussion of fixed and random effects). The posttest scores were entered as the dependent variable with the pretest scores used as a covariate. In other words, a fixedeffects ANCOVA was executed for each independent study. The ANCOVA findings were used to compute observed and corrected effect sizes (Cohen’s d) for each study (see Technical Note 6 for the formula used to compute the effect size). CMA was then used to aggregate the findings from the studies Final Report 11 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 using the observed and corrected effect sizes for the treatment/control condition (i.e., use of Promethean ActivClassroom). The continuation study sought to answer the following evaluation question: Evaluation Question 1: What effect does Promethean ActivClassroom have on students’ achievement regarding the subject matter content taught by their teachers? In addition, the following questions were considered through a meta-analysis of categorical variables: Evaluation Question 2: Does the effect of Promethean ActivClassroom differ between school levels? Evaluation Question 3: Does the effect of Promethean ActivClassroom differ between grade levels? Evaluation Question 4: Does the effect of Promethean ActivClassroom differ between academic content areas? Finally, the following question was considered through a correlational analysis of teacher survey responses: Evaluation Question 5: Is there a relationship between the effect of Promethean ActivClassroom and length of teaching experience, how long teachers have used Promethean ActivClassroom, the percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ self-reported confidence in their use of Promethean ActivClassroom? Findings for each question are discussed separately. Evaluation Question 1: What effect does Promethean ActivClassroom have on students’ achievement regarding the subject matter content taught by their teachers? Figure 18 presents the findings and associated effect size (Cohen’s d) from the ANCOVA for 45 teachers involved in the continuation study. The columns labeled “Adjusted Mean” contain the posttest mean adjusted for differences in the pretest scores for the control and treatment groups respectively (number of students reported in parentheses). The column labeled “ES” contains the calculated effect size (Cohen’s d) for each study (see Technical Note 6), the column labeled “Sig.” contains the p-value (2-tailed) for each study, and the column labeled “% Gain” contains the percentile gain (or loss) associated with each effect size. (For a discussion of effect size and associated percentile gain, see Technical Note 7.) Final Report 12 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 18. ANCOVA Findings for Treatment/Control Studies Site Teacher Grade Adjusted Mean (Control) Adjusted Mean (Treatment) ES 1 1 1 88.03 (n=21) 82.57 (n=20) -.64 .0555 -23.9 2 2 4 78.94 (n=17) 74.29 (n=17) -.30 .4077 -11.8 2 3 3 82.46 (n=17) 95.18 (n=17) 1.03** .0072 34.9 2 4 4 74.71 (n=19) 54.76 (n=15) -1.24** .0017 -39.3 2 5 3 93.33 (n=16) 94.80 (n=16) .19 .6104 7.5 3 6 3 68.16 (n=17) 63.74 (n=14) -.50 .1982 -19.2 4 7 4 47.55 (n=20) 49.00 (n=19) .09 .7889 3.6 5 8a 7 81.16 (n=22) 84.79 (n=26) .31 .3108 12.2 5 8b 8 58.18 (n=20) 67.76 (n=15) .77* .0397 27.9 5 9 8 82.68 (n=21) 92.63 (n=17) .96** .0076 33.2 6 10 9 67.32 (n=25) 64.50 (n=18) -.29 .3732 -11.4 7 11 1 63.70 (n=20) 70.43 (n=19) .45 .1821 17.4 8 12 1 86.99 (n=23) 63.01 (n=23) -1.65**** .0000 -45.1 9 13 8 46.22 (n=9) 83.82 (n=11) 2.54*** .0001 49.5 9 14 8 70.52 (n=17) 82.40 (n=17) .45 .2237 17.4 9 15 7 71.10 (n=14) 85.86 (n=17) 1.81*** .0001 46.5 9 16 8 57.67 (n=13) 63.35 (n=18) .39 .3148 15.2 9 17 7 77.68 (n=10) 81.01 (n=16) .26 .5500 10.3 10 18 9-12 86.94 (n=2) 74.03 (n=4) -1.70 .2600 -45.5 11 19 5 89.94 (n=14) 96.29 (n=17) .61 .1172 22.9 Final Report 13 2009/2010 Second Year Promethean Evaluation Study Sig. (2-tailed) % Gain Marzano Research Laboratory December 2010 Site Teacher Grade Adjusted Mean (Control) Adjusted Mean (Treatment) ES 11 20 5 80.99 (n=16) 78.13 (n=16) -.15 .6983 -6.0 11 21 2 94.25 (n=15) 98.70 (n=16) 1.11** .0067 36.7 11 22 4 89.70 (n=17) 101.75 (n=20) .96** .0087 33.2 11 23 1 69.33 (n=20) 82.43 (n=22) 1.04** .0025 35.1 11 24 2 95.01 (n=15) 95.29 (n=20) .04 .9067 1.6 11 25 1 59.23 (n=18) 61.79 (n=19) .14 .6948 5.6 12 26 K 87.80 (n=12) 93.55 (n=17) .46 .2565 17.7 12 27 5 90.49 (n=13) 94.27 (n=6) .82 .1486 29.4 12 28 5 76.75 (n=17) 85.91 (n=20) 1.21** .0013 38.7 13 29 4 79.59 (n=22) 81.54 (n=24) .22 .4825 8.7 14 30 K 72.31 (n=11) 79.10 (n=16) .59 .1703 22.2 15 31 9-12 81.17 (n=19) 78.24 (n=20) -.26 .4328 -10.6 16 32 4 85.37 (n=19) 92.50 (n=18) .76* .0339 27.6 16 33 1 84.63 (n=14) 95.20 (n=16) 1.70*** .0001 45.5 17 34 7 68.16 (n=22) 77.66 (n=22) .53 .0966 20.2 17 35 6 83.98 (n=20) 88.45 (n=19) .36 .2811 14.4 18 36 5 79.87 (n=24) 93.25 (n=24) 1.50**** .0000 43.3 19 37 4 90.34 (n=12) 88.66 (n=15) -.17 .6810 -6.8 20 38 8 92.32 (n=18) 95.16 (n=21) .17 .6150 6.8 21 39 2 78.48 (n=17) 70.58 (n=17) -.72 .0540 -26.4 Final Report 14 2009/2010 Second Year Promethean Evaluation Study Sig. (2-tailed) % Gain Marzano Research Laboratory December 2010 Site Teacher Grade Adjusted Mean (Control) Adjusted Mean (Treatment) ES 22 40 10-12 66.69 (n=29) 78.39 (n=23) .68* .0215 25.2 23 41 3 72.60 (n=11) 74.02 (n=19) .14 .7352 5.6 24 42 2 58.23 (n=10) 64.76 (n=18) .55 .1966 20.9 25 43 -- 80.33 (n=11) 81.70 (n=12) .13 .7710 5.2 26 44 -- 18.60 (n=18) 18.94 (n=15) .10 .7905 4.0 67.14 (n=12) *p < .05; **p < .01; ***p < .001; ****p < .0001. 65.82 (n=15) -.06 .8762 -2.8 27 45 3 Sig. (2-tailed) % Gain Figure 18 presents the findings for 46 treatment/control studies (note: one of the 45 teachers submitted two studies). When considering the information displayed in this figure, it should be noted that the data for each treatment/control study were checked for obvious coding errors, negative gains, and other potential outliers. A reasonable assumption can be made that students would learn more about academic content addressed during a unit of instruction as the unit progressed. It would not make sense for a student to know less about academic content at the end of a unit. Based on this assumption, students who scored higher on the pretest than the posttest were excluded from analysis. The student counts reflected in this report represent the number of students whose test scores were considered for analysis in the continuation study (i.e., students who took both pretest and posttest that did not score higher on the pretest). The columns labeled “ES” and “% Gain” are of considerable interest. Again, the column entitled “ES” contains the standardized mean difference effect size (Cohen’s d) for each treatment/control study that was computed from the ANCOVA findings for each study. The column entitled “% Gain” contains the percentile gain (or loss) in achievement associated with the treatment (i.e., use of Promethean ActivClassroom). The values in this column were determined by consulting a normal curve table for the area for each reported effect size. A positive value in these columns indicates a finding that favors the treatment group. Consider the results reported for Teacher 3 in the third row of Figure 18. The adjusted mean for the control group was 82.46 (n = 17) and the adjusted mean for the treatment group was 95.18 (n = 17). The percentile gain associated with the effect size (d = 1.03) for this study was 34.9. This means that the average score in the treatment group was 1.03 of a standard deviation or 34.9 percentile points greater than the average score in the control group. In some cases the reported percentile gain was negative. This occurs when the adjusted mean for the treatment group is less than the adjusted mean for the control group. For example, the percentile “gain” reported for Teacher 1 in the first row was negative 23.9 (-23.9, d = -.64). This means that the average score in the treatment group was .64 of a standard deviation or 23.9 percentile points lower than the average score in the control group. Stated differently, Final Report 15 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 the average score in the control group was 23.9 percentile points greater than the average score in the treatment group. (See Technical Note 7 for a more detailed explanation of percentile gain.) In social science research and evaluation it is common practice to consider any contrast with a probability less than .05 as “statistically significant” (see Murphy & Myors, 2004). Using this convention, Figure 18 indicates that the comparison between treatment and control groups was statistically significant for 15 treatment/control studies at the .05 level (p <. 05). Of those 15 studies, 12 would also be considered statistically significant at the .01 level (p < .01), 5 would be considered statistically significant at the .001 level (p < .001), and 2 would be considered statistically significant at the .0001 level. Consider the p-value reported for Teacher 3, p = .0072. This value is generally interpreted to mean that only 7 times in 1,000 would sampling error alone be responsible for the observed difference between group means. The higher this value, the more likely that sampling error alone explains the variance between groups. Again, 15 out of the 46 studies (or 32.6%) were considered statistically significant (p < .05). Conversely, 67.4% were not considered statistically significant. Using meta-analytic techniques, aggregate findings might be considered statistically significant even though a number of the individual studies were not found to be significant at the .05 level (p < .05). Such is the case with the present set of studies. For comparison, Figure 19 shows the overall weighted mean effect size (Cohen’s d) for a meta-analysis of 46 treatment/control studies (continuation study) and 131 treatment/control studies (combined continuation and first year studies) using a random-effects model of error (see Technical Note 8 for discussion of fixed-effect vs. random-effects meta-analysis). The column labeled “N” identifies the ����” reports the weighted number of treatment/control studies included in the group, the column labeled “𝐸𝑆 mean effect size (Cohen’s d) for the studies, the column labeled “SE” contains the standard error for the reported weighted mean effect size, the column labeled “95% CI” identifies the 95% confidence interval (lower limit and upper limit) for the reported weighted mean effect size, the column labeled “Sig.” reports the p-value (2-tailed) for the reported weighted mean effect size, the column labeled “% Gain” contains the percentile gain (or loss) associated with the reported weighted mean effect size, and the column labeled “Fail-Safe N” identifies the number of additional studies with an effect size of .00 that would be required to reduce the weighted mean effect size to .01 using Orwin’s formula (for a discussion of sampling bias and the fail-safe N, see Lipsey & Wilson, 2001, pp. 165-166). Figure 19. Overall Random Effects Evaluation Study N Continuation 46 95% CI % Gain Fail-Safe N .0014 (.0011) 13.3 (15.2) 1,518 (1,748) .36**** .05 .25 .46 .0000 (.41)**** (.06) (.29) (.53) (.0000) Note: Corrected findings are presented in parentheses. ** p < .01; **** p < .0001. 14.1 (15.9) 4,585 (5,240) Combined ���� 𝑬𝑺 .34** (.39)** SE .11 (.12) LL UL .13 (.16) .55 (.62) 131 Sig. (2-tailed) When the results of the 46 treatment/control studies from the continuation study were corrected for attenuation and combined into a weighted average effect size, the corrected effect size was statistically significant at the .01 level (d = .39, p < .01). (For a discussion of combining effect sizes and computing significance levels see Lipsey & Wilson, 2001; for a detailed discussion of attenuation see Hunter & Final Report 16 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Schmidt, 2004.) The overall percentile gain was 15.2. This means that on the average, the use of Promethean ActivClassroom in the continuation study was associated with a gain in student academic achievement of 15.2 percentile-points over what was expected when teachers did not use Promethean ActivClassroom (see Technical Note 7 for a more thorough explanation). Consider the fail-safe N reported in parentheses for the continuation study, 1,748. This indicates that over 1,700 additional treatment/control studies with an effect size of .00 (i.e., no difference between groups) would be needed to reduce the weighted mean effect size of .39 to an effect size of .01. In other words, it would take about 1,750 treatment/control studies with no effect (d = .00) to lower the weighted mean effect size (d = .39) to the smallest positive effect size (d = .01). The column labeled “95% CI” contains the 95% confidence interval for the reported weighted mean effect size. Again, the effect size reported in Figure 19 is a weighted average of all the standardized mean difference effect sizes from the 46 treatment/control studies (see Figure 18). As such, it is considered an estimate of the true effect size of the treatment (i.e., use of Promethean ActivClassroom). This interval includes the range of effect sizes in which one can be 95% certain the true effect size falls. For example, consider the confidence interval reported in parentheses for the continuation study. There is a 95% certainty that the true effect size for the meta-analysis of the 46 treatment/control studies is between the values of .16 and .62. When the confidence interval does not include .00, the weighted mean effect size is determined to be statistically significant at the .05 level (p < .05). In other words, d = .00 would not be considered a reasonable assumption. (For a detailed discussion of the meaning of statistical significance, see Harlow, Muliak, & Steiger, 1997.) These findings are consistent with those reported for the first year evaluation study (see Marzano & Haystead, 2009). When the results of the first year evaluation study and continuation study were corrected for attenuation and combined, the overall percentile gain was 15.9 (d = .41, N = 131, p < .0001). It is common practice when aggregating effect sizes through meta-analysis to perform a test to determine whether those effect sizes are homogeneous. Figure 20 reports the results of the homogeneity analysis for the weighted mean effect sizes listed in Figure 19. In basic terms, homogeneity analysis compares the observed variance in effect sizes to the variance that would be expected from sampling error alone (see Technical Note 9 for a discussion of homogeneity analysis). Figure 20. Homogeneity Analysis Evaluation Study Q Continuation 177.704**** (223.659)**** Sig. (2-tailed) .0000 (.0000) df 45 400.262**** .0000 130 (511.723)**** (.0000) Note: Corrected findings are presented in parentheses. **** p < .0001 Combined Final Report 17 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 All of the Q-values were statistically significant (p < .0001). Therefore, a reasonable assumption can be made that the observed variances in effect sizes is greater than what would likely result from sampling error alone. This suggests that further examination of study characteristics is warranted to determine whether any characteristic is systematically associated with the observed variances. Another way to examine the general effect of the use of Promethean ActivClassroom in the classroom is to consider the distribution of effect sizes. Figure 21 presents the distribution of effect sizes for the continuation study. Figure 22 depicts the distribution of effect sizes for the combined continuation and first year studies. Figure 21. Distribution of Effect Sizes – Continuation Study Figure 21 displays the distribution of “groups” of effect sizes across the 46 treatment/control studies from the continuation study reported in Figure 18 (M = .33, SD = .81, Min = -1.70, Max = 2.54). 3 studies exhibited an effect size less than -1.00 (see first three columns), 9 studies exhibited an effect size between -1.00 and .00 (see fourth through sixth columns), 26 studies exhibited an effect size between .00 and 1.00 (see seventh through ninth columns), and so on. 34 out of the 46 studies (or 73.9%) had a positive effect size. This is very close to the percentage reported (78%) in the first year evaluation study (Marzano & Haystead, 2009). Final Report 18 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 22. Distribution of Effect Sizes – Combined Continuation and First Year Studies Figure 22 displays the distribution of effect sizes across the 131 treatment/control studies from the combined continuation and first year studies (M = .36, SD = .65, Min = -1.70, Max = 2.54). 31 studies exhibited a negative effect (see first through sixth columns), 80 studies exhibited an effect size between .00 and 1.00 (see seventh through tenth columns), 19 studies exhibited an effect size between 1.00 and 2.00 (see eleventh through fourteenth columns), and 1 study exhibited an effect size greater than 2.00 (see last column). 100 out of the 131 studies (or 76.3%) had a positive effect size. The following sections present the findings from a random-effects meta-analysis for each question involving a categorical variable. Each variable was treated as a moderator in the meta-analysis. A moderator is a qualitative or quantitative variable that affects the direction and/or strength of the relation between the dependent and independent variables. As Figures 21 and 22 demonstrate, there was a substantial variation in effect sizes between treatment/control studies. A moderator analysis can help offer insight as to why effect sizes vary from one study to another. The findings will be reported in the same manner as for the overall random effects. Final Report 19 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 2: Does the effect of Promethean ActivClassroom differ between school levels? In order to answer this question, a random-effects meta-analysis was employed using the school level for each treatment/control study as a moderator variable. The meta-analytic findings for this variable are reported in Figures 23 and 24. Figure 23. Random Effects – School Level School Level Elementary School (Grades K-5) Middle School (Grades 6-8) High School (Grades 9-12) Evaluation Study N Continuation 30 .27* (.31)* Combined 66 Continuation 95% CI ���� 𝑬𝑺 SE Sig. (2-tailed) % Gain LL UL .13 (.15) .01 (.02) .52 (.60) .0389 (.0348) 10.6 (12.2) .38**** (.43)**** .08 (.09) .23 (.27) .52 (.60) .0000 (.0000) 14.8 (16.6) 11 .71*** (.82)*** .21 (.24) .29 (.34) 1.13 (1.30) .0009 (.0007) 26.1 (29.4) Combined 43 .33*** (.38)*** .09 (.10) .15 (.18) .50 (.58) .0003 (.0002) 12.9 (14.8) Continuation 5 -.07 (-.10) .33 (.38) -.72 (-.84) .58 (.63) .8232 (.7831) -2.8 (-4.0) .36** .13 .10 .61 .0069 14.1 (.40)** (.15) (.11) (.69) (.0074) (15.5) Note: See discussion of Figure 19 for a description of column headings. Corrected findings are presented in parentheses. * p < .05; ** p < .01; *** p < .001; **** p < .0001. Combined 22 Figure 24. Homogeneity Analysis – School Level Evaluation Study Qb Continuation 4.839 (5.176) Sig. (2-tailed) .0890 (.0752) .184 .9119 (.166) (.9204) Note: Corrected findings are presented in parentheses. Combined df 2 2 Figure 23 depicts the estimates of effect size from a random-effects meta-analysis for the elementary, middle, and high school levels. The weighted mean effect sizes reported for the continuation study were statistically significant for middle school (d = .82, N = 11, p < .001) and elementary school (d = .31, N = 30, p < .05). The weighted mean effect sizes reported for the combined continuation and first year studies were statistically significant for elementary school (d = .43, N = 66, p < .0001), middle school (d = .38, N = 43, p < .001), and high school (d = .40, N = 22, p < .01). The effect sizes were positive for elementary and middle school in the continuation study. When combined with the first year evaluation study, the effect sizes were positive for all three school levels. Final Report 20 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 24 reports the results of the homogeneity analysis for the levels of this moderator variable— school level (see Technical Note 9 for a discussion of homogeneity analysis and moderator variables). Neither the Q-value for the continuation or combined study was statistically significant (p < .05). A reasonable assumption can be made that school level was not a significant contributor to the variation in effect sizes. Evaluation Question 3: Does the effect of Promethean ActivClassroom differ between grade levels? In order to answer this question, a random-effects meta-analysis was employed using the reported grade level for each treatment/control study as a moderator variable. The meta-analytic findings for this variable are reported in Figures 25 and 26. Figure 25. Random Effects – Grade Level Grade Evaluation Study N Continuation 2 .52 (.61) Combined 3 Continuation ���� 𝑬𝑺 95% CI Sig. SE (2-tailed) % Gain LL UL .55 (.62) -.55 (-.61) 1.60 (1.82) .3386 (.3276) 19.9 (22.9) .42 (.48) .39 (.43) -.34 (-.36) 1.78 (1.32) .2763 (.2652) 16.3 (18.4) 6 .15 (.17) .31 (.35) -.46 (-.51) .75 (.86) .6334 (.6190) 6.0 (6.8) Combined 11 .33 (.39) .19 (.21) -.03 (-.02) .70 (.80) .0750 (.0678) 12.9 (15.2) Continuation 4 .23 (.27) .38 (.43) -.52 (-.58) .98 (1.12) .5440 (.5333) 9.1 (10.6) Combined 9 .37 (.43) .21 (.24) -.04 (-.04) .78 (.89) .0802 (.0726) 14.4 (16.6) Continuation 5 .16 (.18) .34 (.39) -.51 (-.57) .83 (.94) .6382 (.6339) 6.4 (7.1) Combined 10 .35 (.40) .20 (.22) -.04 (-.04) .74 (.84) .0757 (.0751) 13.7 (15.5) Continuation 7 .05 (.06) .28 (.32) -.50 (-.57) .61 (.70) .8484 (.8497) 2.0 (2.4) Combined 9 .03 (.04) .21 (.23) -.37 (-.42) .44 (.49) .8692 (.8755) 1.2 (1.6) Continuation 5 .80* (.92)* .35 (.39) .13 (.16) 1.48 (1.69) .0198 (.0181) 28.8 (32.1) Combined 18 .54*** (.62)*** .14 (.16) .26 (.30) .82 (.94) .0002 (.0001) 20.5 (23.2) Continuation 1 .37 (.42) .74 (.85) -1.09 (-1.24) 1.82 (2.08) .6228 (.6186) 14.4 (16.3) Combined 17 .37* (.43)* .15 (.17) .08 (.10) .66 (.76) .0127 (.0116) 14.4 (16.6) K 1 2 3 4 5 6 Final Report 21 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Grade Evaluation Study N Continuation 4 .70 (.81) Combined 14 Continuation 95% CI ���� 𝑬𝑺 Sig. SE (2-tailed) % Gain LL UL .38 (.43) -.05 (-.04) 1.44 (1.65) .0656 (.0614) 25.8 (29.1) .19 (.22) .16 (.18) -.12 (-.14) .50 (.58) .2351 (.2227) 7.5 (8.7) 6 .79* (.92)* .32 (.36) .17 (.22) 1.41 (1.62) .0122 (.0104) 28.5 (32.1) Combined 12 .44* (.52)* .18 (.20) .09 (.12) .79 (.91) .0133 (.0105) 17.0 (19.9) Continuation 1 -.29 (-.33) .74 (.84) -1.73 (-1.98) 1.16 (1.32) .6949 (.6921) -11.4 (-12.9) Combined 7 .33 (.39) .24 (.27) -.13 (-.14) .80 (.91) .1568 (.1462) 12.9 (15.2) Continuation -- -- -- -- -- -- -- Combined -- -- -- -- -- -- -- Continuation -- -- -- -- -- -- -- Combined -- -- -- -- -- -- -- Continuation -- -- -- -- -- -- -- 7 8 9 10 11 12 .13 .69 -1.22 1.49 .8485 5.2 (.15) (.77) (-1.35) (1.65) (.8425) (6.0) Note: See discussion of Figure 19 for a description of column headings. Corrected findings are presented in parentheses. * p < .05; *** p < .001. Combined 1 Figure 26. Homogeneity Analysis – Grade Level Evaluation Study Qb (2-tailed) Continuation 7.248 (7.485) .6113 (.5687) Sig. 5.499 .8554 (5.711) (.8389) Note: Corrected findings are presented in parentheses. Combined df 9 10 Figure 25 displays the random effects for each grade level, K-12. In the continuation study, two studies did not report a grade level and three studies reported more than one grade level. Therefore, five treatment/control studies were excluded from the grade level analysis. The weighted mean effect sizes reported in Figure 25 were statistically significant for two grade levels in the continuation study: fifth grade (d = .92, N = 5, p < .05) and eighth grades (d = .92, N = 6, p < .05). The effect sizes were positive for nine grade levels. When combined with the findings from the first year evaluation study, the Final Report 22 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 weighted mean effect sizes were statistically significant for fifth grade (d = .62, N = 18, p < .001), sixth grade (d = .43, N = 17, p < .05), and eighth grade (d = .52, N = 12, p < .05). The effect sizes were positive for eleven grade levels. Altogether, 20 treatment/control studies were excluded from the combined analysis. Figure 26 depicts the results of the homogeneity analysis for the levels of this moderator variable— grade level. For this moderator variable, the between studies Q-values were not statistically significant (p < .05). Therefore, a reasonable assumption can be made that grade level was not a significant contributor to the variation in effect sizes. Evaluation Question 4: Does the effect of Promethean ActivClassroom differ between academic content areas? In order to answer this question, a random-effects meta-analysis was employed using the reported academic content area (see Appendix C) for each treatment/control study as a moderator variable. The meta-analytic findings for this variable are reported in Figures 27 and 28. Figure 27. Random Effects – Academic Content Area School Level Evaluation Study N Continuation 95% CI ���� 𝑬𝑺 SE 13 .33 (.38) Combined 33 Continuation Sig. (2-tailed) % Gain LL UL .20 (.23) -.07 (-.07) .72 (.82) .1029 (.0959) 12.9 (14.8) .35*** (.40)*** .11 (.12) .14 (.17) .55 (.64) .0009 (.0007) 13.7 (15.5) 18 .23 (.27) .17 (.19) -.09 (-.10) .56 (.64) .1606 (.1533) 9.1 (10.6) Combined 48 .31*** (.36)*** .09 (.10) .14 (.17) .48 (.55) .0003 (.0003) 12.2 (14.1) Continuation 9 .77** (.90)** .24 (.28) .30 (.36) 1.25 (1.44) .0014 (.0011) 27.9 (31.6) Combined 24 .52*** (.60)*** .12 (.14) .28 (.33) .76 (.87) .0000 (.0000) 19.9 (22.6) Continuation 4 .26 (.30) .35 (.40) -.44 (-.49) .95 (1.09) .4657 (.4543) 10.3 (11.8) Language Arts Mathematics Science Social Studies .24 .14 -.02 .51 .0731 9.5 (.28) (.15) (-.02) (.58) (.0655) (11.0) Note: See discussion of Figure 19 for a description of column headings. Corrected findings are presented in parentheses. ** p < .01; *** p < .001. Combined 19 Final Report 23 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 28. Homogeneity Analysis – Academic Content Area Evaluation Study Qb Continuation 3.611 (3.764) Sig. (2-tailed) .3066 (.2881) 2.797 .4239 (2.876) (.4111) Note: Corrected findings are presented in parentheses. Combined df 3 3 Figure 27 lists the meta-analytic findings for the treatment/control studies that involved four academic content areas: language arts, mathematics, science, and social studies. In the continuation study, two studies involved unrelated content areas and were excluded from this meta-analysis. For the continuation study, the weighted mean effect size reported for science was statistically significant (d = .90, N = 9, p < .01). The effect sizes were positive for all four academic content areas. When the findings were combined with the first year evaluation study, the weighted mean effect sizes were statistically significant for language arts (d = .40, N = 33, p < .001), mathematics (d = .36, N = 48, p < .001), and science (d = .60, N = 24, p < .001). The effect sizes were positive for all content areas. Seven treatment/control studies involved unrelated content areas and were excluded from the combined analysis. Figure 28 displays the results of the homogeneity analysis for the levels of this moderator variable— academic content area. For this moderator variable, the between studies Q-values were not statistically significant (p < .05). Therefore, a reasonable assumption can be made that academic content area was not a significant contributor to the variation in effect sizes. Evaluation Question 5: Is there a relationship between the effect of Promethean ActivClassroom and length of teaching experience, how long teachers have used Promethean ActivClassroom, the percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ selfreported confidence in their use of Promethean ActivClassroom? In order to answer this question, SPSS was used to compute the product moment correlation coefficient (Pearson’s r) for a comparison between the observed effect sizes (Cohen’s d) from Figure 18 and the four teacher survey responses (see Appendix C): length of teaching experience (reported in years), how long Promethean ActivClassroom has been used (reported in months), percentage of time Promethean ActivClassroom was used in the classroom, and teachers’ confidence in their ability to use Promethean ActivClassroom. For comparison purposes, Figures 29 through 36 display the distribution for each of the survey responses. Final Report 24 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 29. Distribution of Teaching Experience – Continuation Study Figure 29 shows that for the continuation study, 21 of the treatment/control studies were conducted by teachers with less than 10 years of experience, 17 studies were conducted by teachers with between 10 and 20 years of experience, 6 studies were conducted by teachers with between 20 and 30 years of experience, and 1 study was conducted by a teacher with more than 30 years of experience. (Note: survey data missing for one study.) Final Report 25 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 30. Distribution of Teaching Experience – Combined Continuation and First Year Studies Figure 30 indicates that for the combined continuation and first year studies, 55 of the treatment/control studies were conducted by teachers with less than 10 years of experience, 44 studies were conducted by teachers with between 10 and 20 years of experience, 14 studies were conducted by teachers with between 20 and 30 years of experience, and 3 studies were conducted by teachers with more than 30 years of experience. (Note: survey data missing for 15 studies.) Final Report 26 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 31. Distribution of Promethean ActivClassroom Use – Continuation Study Figure 31 shows that for the continuation study, 15 of the treatment/control studies were conducted by teachers who reported having used Promethean ActivClassroom for less than 10 months, 14 studies were conducted by teachers who reported having used Promethean ActivClassroom between 10 and 20 months, 4 studies were conducted by teachers who reported having used Promethean ActivClassroom between 20 and 30 months, 9 studies were conducted by teachers who reported having used Promethean ActivClassroom between 30 and 40 months, and 2 studies were conducted by teachers who reported having used Promethean ActivClassroom for more than 40 months. (Note: survey data missing for two studies.) Final Report 27 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 32. Distribution of Promethean ActivClassroom Use – Combined Continuation and First Year Studies Figure 32 indicates that for the combined continuation and first year studies, 38 of the treatment/control studies were conducted by teachers who reported having used Promethean ActivClassroom for less than 10 months, 28 studies were conducted by teachers who reported having used Promethean ActivClassroom between 10 and 20 months, 20 studies were conducted by teachers who reported having used Promethean ActivClassroom between 20 and 30 months, 21 studies were conducted by teachers who reported having used Promethean ActivClassroom between 30 and 40 months, and 8 studies were conducted by teachers who reported having used Promethean ActivClassroom for more than 40 months. (Note: survey data missing for 16 studies.) Final Report 28 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 33. Distribution of Percentage of Class Time – Continuation Study Figure 33 indicates that for the continuation study, 4 of the treatment/control studies were conducted by teachers who reported having used Promethean ActivClassroom between 20 and 40 percent of their classroom instructional time, 12 studies were conducted by teachers who reported having used Promethean ActivClassroom between 40 and 60 percent of their classroom instructional time, 8 studies were conducted by teachers who reported having used Promethean ActivClassroom between 60 and 80 percent of their classroom instructional time, and 19 studies were conducted by teachers who reported having used Promethean ActivClassroom between 80 and 100 percent of their classroom instructional time. (Note: survey data missing for three studies.) Final Report 29 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 34. Distribution of Percentage of Class Time – Combined Continuation and First Year Studies Figure 34 shows that for the combined continuation and first year studies, one treatment/control study was conducted by a teacher who reported having used Promethean ActivClassroom less than 20 percent of his/her classroom instructional time, 15 studies were conducted by teachers who reported having used Promethean ActivClassroom between 20 and 40 percent of their classroom instructional time, 24 studies were conducted by teachers who reported having used Promethean ActivClassroom between 40 and 60 percent of their classroom instructional time, 28 studies were conducted by teachers who reported having used Promethean ActivClassroom between 60 and 80 percent of their classroom instructional time, and 46 studies were conducted by teachers who reported having used Promethean ActivClassroom between 80 and 100 percent of their classroom instructional time. (Note: survey data missing for 17 studies.) Final Report 30 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 35. Distribution of Confidence Ratings – Continuation Study Final Report 31 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 36. Distribution of Confidence Ratings – Combined Continuation and First Year Studies Figure 35 shows that a majority of the teachers in the continuation study rated themselves as mostly confident in their ability to use Promethean ActivClassroom. (Note: survey data missing for one study.) Figure 36 indicates a similar majority for the combined continuation and first year studies. (Note: survey data missing for 15 studies.) Figures 29 through 36 provide a graphical depiction of the four survey responses. Figure 37 lists the product moment correlation coefficients (Pearson’s r) between the observed effect sizes (Cohen’s d) from the continuation study reported in Figure 18 and the four survey responses. Figure 38 displays the correlation coefficients (Pearson’s r) between the observed effect sizes (Cohen’s d) from the combined continuation and first year studies and the four survey responses. Final Report 32 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 37. Product Moment Correlations for Teacher Survey Responses – Continuation Study Comparison Correlation Effect Size Years of Teaching Experience Months of Technology Use Percentage of Class Time Confidence Effect Size # Yrs Teach #Mos Tech % Class Time Confidence 1 -.09 -.21 .22 -.08 .545 .168 .155 .581 Sig. (2-tailed) N 46 45 44 43 45 Correlation -.09 1 .15 -.18 -.02 Sig. (2-tailed) .545 .347 .248 .912 N 45 45 44 43 45 Correlation -.21 .15 1 .13 .57*** Sig. (2-tailed) .168 .347 .409 .000 N 44 44 44 42 44 Correlation .22 -.18 .13 1 .19 Sig. (2-tailed) .155 .248 .409 N 43 43 42 43 43 Correlation -.08 -.02 .57*** .19 1 Sig. (2-tailed) .581 .912 .000 .222 N 45 45 44 43 .222 45 *** p < .001. Figure 37 shows that no statistically significant correlations (p < .05) were found between the observed effect sizes and the teacher survey responses. However, the correlation between the number of months Promethean ActivClassroom was used and teachers’ confidence in their ability to use Promethean ActivClassroom was positive and significant (r = .57, N = 44, p < .001). Final Report 33 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 38. Product Moment Correlations for Teacher Survey Responses – Combined Continuation and First Year Studies Comparison Correlation Effect Size Years of Teaching Experience Months of Technology Use Percentage of Class Time Confidence Effect Size Yrs Teach Mos Tech % Class Time Confidence 1 -.07 -.03 .22* .04 .490 .725 .022 .693 Sig. (2-tailed) N 131 116 115 114 116 Correlation -.07 1 .01 -.18 -.21* Sig. (2-tailed) .490 .910 .064 .026 N 116 116 115 113 116 Correlation -.03 .01 1 .11 .50*** Sig. (2-tailed) .725 .910 .272 .000 N 115 115 115 112 115 Correlation .22* -.18 .11 1 .45*** Sig. (2-tailed) .022 .064 .272 N 114 113 112 114 113 Correlation .04 -.21* .50*** .45*** 1 Sig. (2-tailed) .693 .026 .000 .000 N 116 116 115 113 .000 116 * p < .05; *** p < .001. Figure 38 indicates that the correlation between the observed effect sizes (combined continuation and first year studies) and the percentage of time Promethean ActivClassroom was used in the classroom was positive and significant (r = .22, N = 114, p < .05). In addition, statistically significant correlations were found between teachers’ confidence and the number of years of teaching experience (r = -.21, N = 116, p < .05), the number of months Promethean ActivClassroom was used (r = .50, N = 115, p < .001), and the percentage of time Promethean ActivClassroom was used (r = .45, N = 113, p < .001). Final Report 34 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Phase I Summary and Interpretation The results of the continuation study are rather straight-forward and might be summarized by the following statements: • • • • When the effect sizes from the 46 treatment/control studies were corrected for attenuation due to unreliability of the dependent measure, the overall effect for the use of Promethean ActivClassroom was .39 and was significant at the .01 level (p < .01). No significant relationship was found between the 46 effect sizes associated with the use of Promethean ActivClassroom and the following teacher survey response variables: o length of teaching experience (reported in years) o how long teachers have used Promethean ActivClassroom (reported in months) o the percentage of time Promethean ActivClassroom was used in the classroom o teachers’ self-reported confidence in their use of Promethean ActivClassroom A significant relationship was found between teachers’ self-reported confidence in their ability to use Promethean ActivClassroom and the number of months Promethean ActivClassroom was used (r = .57, N = 44, p < .001). No significant mediator influence was found for the effects of Promethean ActivClassroom for any of the following variables: school level, grade level, and academic content area. Perhaps even more important than the findings from the continuation study are the combined findings from the first year and continuation studies. These combined findings can be summarized in the following statements: • When the effect sizes from the 131 treatment/control studies were corrected for attenuation due to unreliability of the dependent measure, the overall effect for the use of Promethean ActivClassroom was .41 and was significant at the .0001 level (p < .0001). This statement implies that the observed differences between treatment and control groups could have occurred less than one time in ten thousand if there was no true relationship between Promethean ActivClassroom and student academic achievement. A reasonable inference is that the 15.9 percentilepoint gain associated with the overall weighted average effect size of .41 is probably not a function of random factors that are specific to the treatment/control studies. Stated differently, a reasonable assumption is that the use of Promethean ActivClassroom is associated with a gain in student achievement. • A significant relationship (r = .22, N = 114, p < .05) was found between the 131 effect sizes associated with the use of Promethean ActivClassroom and the percentage of time Promethean ActivClassroom was used in the classroom. This statement implies that the effect of Promethean ActivClassroom on student academic achievement increases as the percentage of classroom instructional time that Promethean ActivClassroom is utilized increases. Marzano and Haystead (2009) reported findings that suggested substantial gains in student achievement under the following conditions: a teacher has 10 years or more of teaching experience, a teacher has used Promethean ActivClassroom for two years or more, a teacher uses Promethean ActivClassroom between 75 and 80 percent of the time in his or her classroom, and a teacher has high Final Report 35 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 confidence in his or her ability to use Promethean ActivClassroom. The correlations reported in Figure 38 indicate that except for the percentage of class time in which Promethean ActivClassroom was used, those conditions do not have a relationship with effect size. Relative to the percentage of time Promethean ActivClassroom was used, the positive relationship with effect size does not decrease after 80 percent of class time. • Significant relationships were found between teachers’ self-reported confidence in their ability to use Promethean ActivClassroom and the following teacher survey response variables: o length of teaching experience [reported in years] (r = -.21, N = 116, p < .05) o how long teachers have used Promethean ActivClassroom [reported in months] (r = .50, N = 115, p < .001) o the percentage of time Promethean ActivClassroom was used in the classroom (r = .45, N = 113, p < .001) This statement implies that teachers’ confidence ratings were influenced by how long they have been a classroom teacher and how long they have used Promethean ActivClassroom. Additionally, this statement implies that the percentage of time Promethean ActivClassroom was used in the classroom was influenced by teachers’ confidence ratings. The negative correlation between confidence and teaching experience suggests that as the length of teaching experience increased confidence ratings decreased. In other words, some of the teachers in the evaluation study with several years of experience in the classroom were less confident in their ability to use Promethean ActivClassroom. Conversely, some of the teachers with a few years of experience in the classroom were more confident in their ability to use Promethean ActivClassroom. Again, 99 of the 131 effect sizes (or 75.6%) were contributed by teachers with less than 20 years of teaching experience (see Figure 30). The most parsimonious explanation for the negative relationship between confidence and the number of years of teaching experience might simply be the lack of effect sizes contributed by teachers with more than 20 years of teaching experience. Should additional treatment/control studies be conducted by teachers with more than 20 years of experience, the finding reported here would most likely change. The positive correlation between confidence and the number of months Promethean ActivClassroom was used suggests that as the use of Promethean ActivClassroom increased confidence ratings increased. Stated differently, teachers in the evaluation study who had used Promethean ActivClassroom for the longest period of time were more confident in their ability to use Promethean ActivClassroom. A reasonable inference can be made that increased confidence will result from increased use of Promethean ActivClassroom over a period of time. The positive correlation between confidence and the percentage of time Promethean ActivClassroom was used in the classroom suggests that the amount of instructional time Promethean ActivClassroom was used increased as confidence ratings increased. In other words, teachers in the evaluation study with higher confidence ratings used Promethean ActivClassroom more often during instructional time. This implies that increased use of Promethean ActivClassroom will result from increased confidence in the ability to use Promethean ActivClassroom. Additionally, given that there was a positive correlation between the percentage of time Promethean ActivClassroom was used in the classroom and effect size, a reasonable inference can be made that teachers’ confidence has an indirect relationship with effect size. Final Report 36 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 • No significant mediator influence was found for the effects of Promethean ActivClassroom for any of the following variables: school level, grade level, and academic content area. This statement implies that school level, grade level, and academic content area were not significant contributors to the variance between the 131 effect sizes. Some other aspect (or aspects) of the treatment/control studies may have contributed to that variance. The findings for these three variables suggest that Promethean ActivClassroom can be equally effective at all school levels, grade levels, and academic content areas. Marzano and Haystead (2009) reported lower than anticipated effects at seventh grade. The random effects reported in Figure 25 indicate that additional studies did improve the average effect for this grade level. Given the significant findings from the homogeneity analysis reported in Figure 20 the most likely explanation for the variance in effect sizes would be characteristics associated with the classroom teacher not already considered in this report. Final Report 37 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Phase II Final Report 38 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Phase II involved an analysis of student learning with and without Promethean ActivClassroom from the perspective of teacher behavior as evidenced in videotapes of teachers using Promethean ActivClassroom tools in their classrooms. Teachers in the study were asked to be videotaped for at least one lesson in which they used Promethean ActivClassroom (i.e., treatment group) and one lesson in which they did not (i.e., control group). This allowed for the examination of a number of evaluation questions that were not available without the videotapes. The general focus of Phase II was to determine the behaviors that differentiated those teachers who obtained positive effects from Promethean ActivClassroom from those who did not. Recall that in the Phase I analysis 74% of the treatment/control studies in the second year evaluation study exhibited greater achievement with Promethean ActivClassroom and 26% of the treatment/control studies exhibited greater achievement without the technology. The evaluation questions that relate to this phase of the study were: Evaluation Question 1: What is the relationship between student engagement and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 2: What is the relationship between teacher IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 3: What is the relationship between student IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 4: What is the relationship between multiple student use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 5: What is the relationship between student independent use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 6: What is the relationship between teacher use of IWB reinforcers and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 7: What is the relationship between teacher use of learner response systems and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 8: What is the relationship between teacher use of the IWB to represent knowledge graphically or nonlinguistically and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 9: What is the relationship between teacher previewing of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 10: What is the relationship between teacher chunking of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 11: What is the relationship between teacher scaffolding of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 12: What is the relationship between teacher pacing of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 13: What is the relationship between teacher monitoring of student progress and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 14: What is the relationship between clarity of content as depicted in the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 15: What is the relationship between student interaction about the content and positive versus negative effects for the use of Promethean ActivClassroom? Final Report 39 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 16: What is the relationship between student response rates and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 17: What is the relationship between classroom management and positive versus negative effects for the use of Promethean ActivClassroom? In general, two types of analyses were performed for each of these questions. One analysis broke the teachers into two categories: those with positive effects and those with negative effects for the use of Promethean ActivClassroom. Membership in these groups was considered the independent variable and scores on the variables for the 17 Phase II evaluation questions were considered the dependent variables. SPSS was used to perform an independent-samples t-test comparing the mean for the group with positive effects and the mean for the group with negative effects for each of the 17 Phase II variables. SPSS generates output for two t-tests based on assumptions of homogeneous variance: equal variances assumed and equal variances not assumed. Unless otherwise noted, the t-value reported for each variable was based on the t-test with equal variances assumed. (See Technical Note 10 for a discussion of the ttest and methods for assessing homogeneous variance.) For the second analysis SPSS was used to determine the bivariate correlation (Pearson’s r) between the scores on the 17 Phase II variables and the corrected effect sizes computed in Phase I. (See Technical Note 4 for the method that was utilized to correct for attenuation in the observed effect sizes reported in Figure 18.) Coding of Videotapes Each videotape was coded by a single rater. Intra-rater reliability was computed by coding 10 videos on all 17 variables and then re-coding those videos two weeks later without access to previous ratings. Agreement between the two ratings was determined by analyzing the percentage of scores from the second rating that were: 1) the same as the first rating and 2) one point (plus or minus) from the first rating. These percentages are reported in Figure 39. Figure 39. Percentage of Second Ratings the Same as First Rating and One Score Point from First Rating Variable Same Rating One Point Difference Combined 1 60% 20% 80% 2 70% 20% 90% 3 80% 10% 90% 4 60% 10% 70% 5 60% 20% 80% 6 60% 10% 70% 7 70% 10% 80% 8 60% 20% 80% 9 70% 20% 90% Final Report 40 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Variable Same Rating One Point Difference Combined 10 60% 20% 80% 11 60% 40% 100% 12 60% 20% 80% 13 70% 20% 90% 14 60% 20% 80% 15 70% 20% 90% 16 60% 30% 90% 17 60% 40% 100% As indicated in the last column of Figure 39, on 15 out of 17 Phase II variables the second rating was either identical to the first or different by one point only (plus or minus) 80% of the time or more. Evaluation Question 1: What is the relationship between student engagement and positive versus negative effects for the use of Promethean ActivClassroom? Engagement refers to the extent to which students were paying attention to the classroom activities. High scores on this variable are exhibited by students consistently attending to the activities orchestrated by the teacher. Figure 40 reports the results for the comparison of means for student engagement between those teachers who exhibited positive effects for the use of Promethean ActivClassroom and those students who exhibited negative effects for the technology. Figure 40. Comparison of Means for Student Engagement Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 3 7 7 2 1 0 0 5.45 (N=20) .23 Negative 1 0 3 4 0 0 0 4.75 (N=8) .37 Positive 26 16 19 14 2 0 0 5.65 (N=77) .14 0 4.71 (N=24) .22 Continuation Combined Negative 2 4 4 13 1 0 t Sig. (2-tailed) df 1.60 .122 26 3.43** .001 99 ** p < .01. Final Report 41 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 As shown in Figure 40, the t-test reveals that in the continuation study, the positive-effects group exhibited a higher mean score for engagement than the negative-effects group. The difference between group means was not statistically significant at the .05 level (p < .05). When combined with the findings from the first year evaluation study, the t-test indicates a significant difference (p < .01) between mean scores favoring the positive-effects group. One thing to note regarding these findings is that both mean scores were above the middle value on the seven-point scale that was used (highest score = 7; lowest score = 1). Taken at face value, the mean scores suggest a high central tendency for both groups on that scale. However, mean scores can sometimes obscure characteristics of data points within a sample, especially when the sample size is small. For example, a few data points at the high end of a scale can inflate the mean score for that data set even though a majority of the data points are at the lower end of the scale. Therefore, it is often useful to compare a distribution of frequencies to determine whether the true central tendency lies at, above, or below the mean score for the distribution. Consider the findings reported for the continuation study. In the positive-effects group, the median (i.e., middle value of the data set) was 5.5 and there were two modes (i.e., most frequent value of the data set), 5 and 6. This can be interpreted to mean that the central tendency for engagement was higher than the mean score of 5.45 for this group. Therefore, the central tendency for the positive-effects group was very high. In the negative-effects group, the median was 4.5 and the mode was 4. This suggests that the central tendency for engagement was lower than the mean score of 4.75 for this group. Therefore, the central tendency for the negative-effects group was closer to the middle value on the seven-point scale. When combined with the findings from the first year evaluation study, the median in the positive-effects group was 6 and the mode was 7. Again, the measures of central tendency were very high for this group. In the negative-effects group, the median and mode were both 4. This would seem to indicate a central tendency toward the middle of the scale for this group. Taken at face value, these findings suggest that student engagement was very high in the positive-effects group. In the negative-effects group, student engagement appears to be more moderate. Figure 41 reports the bivariate correlation between student engagement and the corrected effect sizes. Technical Note 11 explains why only the corrected effect sizes were used when computing bivariate correlations for Evaluation Questions 1 through 17 of Phase II. Figure 41. Bivariate Correlation for Student Engagement Pairs Engagement & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .34 (N=28) .079 Combined .44*** (N=101) .000 (2-tailed) *** p < .001. Final Report 42 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 In the continuation study, the bivariate correlation between student engagement and the corrected effect sizes was positive. While the correlation was not significant at the .05 level (p < .05), the p-value was less than .10. Because the correlation approached significance, a reasonable inference can be made that a moderate positive linear relationship was found between effect size and student engagement. This inference is strengthened by the significant correlation (p < .001) reported for the combined continuation and first year studies. Evaluation Question 2: What is the relationship between teacher IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Teacher IWB skills refers to the teacher’s use of features on the IWB. High scores on this variable are exhibited by teachers utilizing various Promethean ActivClassroom tools with ease and fluency. Figure 42 reports the comparison of mean scores on this variable for the positive- and negative-effects classes. Figure 42. Comparison of Means for Teacher IWB Skills Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 3 5 6 2 0 1 0 5.35 (N=17) .31 Negative 0 1 2 2 1 1 0 4.14 (N=7) .51 Positive 24 8 14 18 1 9 0 5.12 (N=74) .20 2 4.09 (N=23) .37 Continuation Combined Negative 3 2 3 8 2 3 t Sig. (2-tailed) df 2.09* .049 22 2.54* .013 95 * p < .05. The t-test reported in Figure 42 shows that the positive-effects group had a higher mean score with respect to teacher IWB skills in the continuation study. The difference between group means was statistically significant at the .05 level (p < .05). When the findings were combined with the first year evaluation study, a significant difference (p < .05) between means was found that favors the positiveeffects group. In the continuation study, both group means were above the middle value on the scale. In the positiveeffects group, the median and mode were both 5. In the negative-effects group, the data set had two modes (4 and 5) with a median of 4. For both groups, the median values were lower than their respective means. In the combined continuation and first year studies, the median in the positive-effects group was 5 and the mode was 7. In the negative-effects group, the median and mode were 4. This suggests a high central tendency for the positive-effects group and a medium central tendency for the negative-effects group. In other words, teachers in the positive-effects group appeared to utilize Promethean ActivClassroom tools with ease and fluency more often than teachers in the negative-effects group. Final Report 43 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 43 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 43. Bivariate Correlation for Teacher IWB Skills Pairs Teacher IWB Skills & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .59** (N=24) .003 Combined .38*** (N=97) .000 (2-tailed) ** p < .01; *** p < .001. Figure 43 reveals a significant correlation (p < .01) between teacher IWB skills and effect size in the continuation study. This finding suggests that a strong positive linear relationship was found between effect size and teachers’ ability to utilize various Promethean ActivClassroom tools with ease and fluency. When combined with the first year evaluation study, the correlation was reduced from .59 to .38 and was significant at the .001 level (p < .001). The combined data would seem to suggest a moderate positive linear relationship between effect size and teachers’ skills with Promethean ActivClassroom tools; as one increases so does the other. Evaluation Question 3: What is the relationship between student IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Student IWB skills refers to the student’s ability to use features of the IWB without direct prompting from the teacher. High scores on this variable manifest as students using features such as highlighting or erasing without need of teacher direction. Figure 44 reports the mean scores on this variable for the positive- and negative-effects classes. Figure 44. Comparison of Means for Student IWB Skills Evaluation Study Frequencies T-Test for Equality of Means Effects Positive Mean SE 7 6 5 4 3 2 1 1 1 2 5 0 1 0 4.50 (N=10) .67 Continuation t df .43 Negative 0 0 0 2 0 1 0 3.33 (N=3) Positive 3 1 6 28 4 7 18 3.18 (N=67) .21 Negative 0 1 3 8 1 1 5 3.32 (N=19) .38 Combined Sig. (2-tailed) 1.34 .206 11 -.32 .753 84 The t-test indicates that the positive-effects group had a higher mean score than the negative-effects group in the continuation study. The difference in means between the two groups was not significant (p < .05). When combined with the first year evaluation study, the negative-effects group exhibited a higher mean score. The difference in means was not significant at the .05 level (p < .05). Final Report 44 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 In the continuation study, the mean for the positive-effects group was above the middle value on the scale. In the positive-effects group, the median and mode were 4. In the negative-effects group, the median and mode were also 4. Therefore, the measures of central tendency in both groups were at the middle of the scale for this variable. In the combined continuation and first year studies, the means for both groups were below the middle value on the scale. The median and mode were 4 in both groups. This suggests a central tendency at the middle of the scale. There appears to be very little difference between groups with respect to observations of students’ ability to use features of Promethean ActivClassroom without direct prompting from the teacher. Figure 45 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 45. Bivariate Correlation for Student IWB Skills Pairs Student IWB Skills & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .60* (N=13) .030 Combined .07 (N=86) .530 (2-tailed) * p < .05. Figure 45 shows that in the continuation study a significant correlation (p < .05) exists between student IWB skills and effect size. This suggests that a strong positive linear relationship was found between effect size and students’ ability to use features of the IWB without prompting from the teacher. However, when combined with the first year evaluation study, the correlation was reduced from .60 to .07. The combined finding was not significant at the .05 level (p < .05). It should be noted that in the continuation study, the correlation was statistically significant but the difference between means was not. This is not uncommon when a variable which is inherently ordinal, interval, or ratio in nature is dichotomized to artificially create two groups for the purpose of comparing differences on other variables. In this case the variable of student IWB skills is ordinal in nature with seven possible values—1 to 7. The effect size variable is a ratio scale. In the analysis reported in Figure 44, the effect size variable was dichotomized (i.e., positive effects and negative effects) in order to make a comparison of means on the variable of student IWB skills. This loss of information due to collapsing of values renders the comparison between groups reported in Figure 44 less representative of the true relationship between variables than the correlations reported in Figure 45. Taken at face value, the findings from the combined continuation and first year studies appear to indicate no linear relationship between effect size and student IWB skills. Final Report 45 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 4: What is the relationship between multiple student use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Multiple student use of the IWB refers to the extent that multiple students use the IWB under the teacher’s direction. High scores on this variable manifest as many or all students in class coming to the front of the room and using the IWB in response to directions from the teacher. Figure 46 reports the difference in means between the positive- and negative-effects classes. Figure 46. Comparison of Means for Multiple Student Use of the IWB Evaluation Study Frequencies T-Test for Equality of Means Effects Positive Mean SE 7 6 5 4 3 2 1 0 3 3 6 1 3 4 3.50 (N=20) .39 .31 Continuation t Negative 0 0 0 0 2 2 4 1.75 (N=8) Positive 8 7 8 21 6 5 22 3.53 (N=77) .23 13 2.21 (N=24) .32 Combined Negative 0 2 1 3 0 5 Sig. (2-tailed) df 3.48** .002 24.28 3.32** .002 48.76 ** p < .01. In the continuation study, the difference in means between the two groups was statistically significant at the .01 level (p < .01). It should be noted that Levene’s test for equality of variances was significant (F = 5.57, p = .026). When combined with the first year findings, the difference in means was also significant at the .01 level (p < .01). Levene’s test was also significant (F = 4.50, p = .036). In both the continuation and combined studies, the assumption that the two groups have approximately equal variance on the dependent variable was not met. Therefore, the t-test for unequal variances was employed (see Technical Note 10). In the continuation study, both means were below the middle value on the scale. In the positive-effects group, the median and mode were 4. In the negative-effects group, the median was 1.5 and the mode was 1. For the positive-effects group, the median value was higher than the mean score. For the negative-effects group, the median was lower than the mean. The measures of central tendency for this variable were low for the negative-effects group. This indicates that observed classes exhibited very little multiple student use of the IWB. When combined with the first year findings, both means were also below the middle value of the scale. In the positive-effects group, the median was 4 and the mode was 1. In the negative-effects group, the median was 2 and the mode was 1. The combined findings suggest a low central tendency for both groups on this variable. In other words, with respect to multiple student use of the IWB, the majority of observations were at the low end of the scale. Figure 47 reports the bivariate correlation between this variable and the corrected effect sizes. Final Report 46 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 47. Bivariate Correlation for Multiple Student Use of the IWB Pairs Multiple Student Use & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .51** (N=28) .006 Combined .44*** (N=101) .000 (2-tailed) ** p < .01; *** p < .001 . Figure 47 indicates that a significant correlation (p < .01) exists between multiple student use of the IWB and effect size in the continuation study. This finding would suggest an increase in effect size when there is an increase in the utilization of the IWB by multiple students in response to teacher directions. The combined finding reveals a significant correlation (p < .001) between multiple student use of the IWB and effect size. A reasonable inference can be made that a strong positive linear relationship was found between effect size and multiple student use of the IWB. Evaluation Question 5: What is the relationship between student independent use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Student independent use of the IWB refers to the extent to which students used the technology without the direct guidance of the teacher. High scores on this variable are exhibited by students working with the IWB in ways that are not directly guided by the teacher. Figure 48 reports the difference in means between the positive-effects class and the negative-effects class for this variable. Figure 48. Comparison of Means for Student Independent Use of the IWB Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 1 0 0 0 0 0 0 7.00 (N=1) -- Negative 0 0 0 0 0 0 1 1.00 (N=1) -- Positive 2 0 0 4 2 2 48 1.52 (N=58) .18 13 1.47 (N=17) .24 Continuation Combined Negative 0 0 0 2 0 2 t Sig. (2-tailed) df -- -- 0 .13 .895 73 For this variable, there were only two treatment/control studies in the continuation study which exhibited student independent use of the IWB, one in each group. In the positive-effects group, student independent use was observed at the highest end of the scale. In the negative-effects group, student independent use was observed at the lowest end of the scale. Therefore, no comparison of means was possible with respect to student independent use of the IWB. When combined with the findings from the first year evaluation study, the mean score was slightly higher in the positive-effects group. The difference between means was not significant at the .05 level (p < .05). Final Report 47 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 The combined findings reveal that both means were below the middle value of the scale. In the positiveeffects group, the median and mode were both 1. The median and mode were the same in the negativeeffects group. The combined findings suggest a very low central tendency for both groups on this variable. In other words, the vast majority of observations of student independent use of Promethean ActivClassroom were at the lower end of the scale. Figure 49 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 49. Bivariate Correlation for Student Independent Use of the IWB Pairs Student Independent Use & Corrected Effect Size Sig. Evaluation Study Correlation Continuation 1.00 (N=2) -- Combined .09 (N=75) .452 (2-tailed) Again, there were only two treatment/control studies in the continuation study which exhibited student independent use of the IWB, one in each group. Because the two observations were at the opposite ends of the scale, the correlation was 1.00. The combined finding reveals a small positive correlation between student independent use and effect size which was not significant at the .05 level (p < .05). This would suggest no linear relationship between effect size and student independent use of the IWB. Evaluation Question 6: What is the relationship between teacher use of IWB reinforcers and positive versus negative effects for the use of Promethean ActivClassroom? IWB reinforcers refer to the use of those Promethean ActivClassroom tools that help indicate when a correct answer has been produced or identified by students. High scores on this variable manifest as frequent use of specific tools (e.g. drop and drag, applause when correct answer is identified or produced). Figure 50 reports the difference in means between the positive-effects and negative-effects classes for this variable. Figure 50. Comparison of Means for Use of IWB Reinforcers Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 2 2 0 4 0 4 5 3.24 (N=17) .53 Negative 0 1 0 0 0 3 4 2.00 (N=8) .60 Positive 15 4 2 15 2 6 30 3.34 (N=74) .28 13 2.17 (N=24) .34 Continuation Combined Negative 0 2 1 3 0 5 t Sig. (2-tailed) df 1.40 .174 23 2.65* .011 55.33 * p < .05. Final Report 48 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 The t-test for teacher use of IWB reinforcers in the continuation study shows that the positive-effects group had a higher mean score than the negative-effects group. The difference in means between the two groups was not significant (p < .05). The combined findings indicate a significant difference (p < .05) between group means favoring the positive-effects group. The t-test for unequal variances was employed because Levene’s test was significant (F = 10.71, p = .001). The mean scores for both groups in the continuation study were below the middle value on the scale. In the positive-effects group, the median was 2. In the negative-effects group, the median was 1.5. In both groups, the mode was 1. Therefore, the measures of central tendency in both groups were at the low end of the scale for this variable. When combined with the findings from the first year evaluation study, the mean scores for both groups were also below the middle value on the scale. In the positive-effects group, the median was 3 and the mode was 1. In the negative-effects group, the median and mode were both 1. For both groups, the measures of central tendency were at the low end of the scale. This suggests that most of the observations of teachers’ utilization of IWB reinforcers were lower than the middle value of the scale. Figure 51 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 51. Bivariate Correlation for Teacher Use of IWB Reinforcers Pairs IWB Reinforcers & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .45* (N=25) .025 Combined .40*** (N=98) .000 (2-tailed) * p < .05; *** p < .001. Figure 51 depicts a significant correlation (p < .05) between teacher use of IWB reinforcers and effect size in the continuation study. This finding implies that as teachers’ utilization of tools that help to indicate students’ correct answers increases, effect size tends to increase as well. The combined finding also reveals a significant correlation (p < .001) suggesting that a strong positive linear relationship was found between effect size and teachers’ utilization of IWB reinforcers. Evaluation Question 7: What is the relationship between teacher use of learner response systems and positive versus negative effects for the use of Promethean ActivClassroom? Teacher use of learner response systems indicates the extent to which teachers have students answer questions using one of Promethean’s learner response systems (ActiVote or ActivExpression). High scores on this variable are indicated when teachers present students with questions and have each student vote on the questions. Figure 52 reports the difference in means between the positive- and negative-effects classes. Final Report 49 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 52. Comparison of Means for Use of Learner Response Systems Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 1 3 3 5 0 6 2 3.70 (N=20) .41 Negative 0 0 0 0 0 5 3 1.63 (N=8) .18 Positive 18 4 3 13 0 7 32 3.42 (N=77) .28 12 2.29 (N=24) .37 Continuation Combined Negative 2 0 0 4 1 5 t Sig. (2-tailed) df 4.61*** .000 24.65 2.40* .020 51.96 * p < .05; *** p < .001. In the continuation study, a significant difference (p < .001) was found between group means favoring the positive-effects group. Levene’s test for equality of variances was significant (F = 11.31, p = .002). When combined with the findings from the first year, a significant difference (p < .05) was also found which favored the positive-effects group. As was the case for the continuation study, Levene’s test was significant (F = 12.56, p = .001). Therefore, the t-test for unequal variances was performed for both the continuation and combined studies. The mean scores for both groups in the continuation study were below the middle value on the scale. In the positive-effects group, the median was 4 and the mode was 2. Two of the three measures of central tendency were near the middle of the scale for this variable. In the negative-effects group, the median and mode were 2. Therefore, the measures of central tendency were at the low end of the scale. The combined findings show that the mean scores were also below the middle value on the scale. In the positive-effects group, the median was 2 and the mode was 1. In the negative-effects group, the median was 1.5 and the mode was 1. The measures of central tendency for both groups were at the low end of the scale. This seems to suggest that a majority of the observations of teachers’ utilization of learner response systems were lower than the middle value of the scale. Figure 53 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 53. Bivariate Correlation for Use of Learner Response Systems Pairs Learner Response Systems & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .49** (N=28) .008 Combined .35*** (N=101) .000 (2-tailed) ** p < .01; *** p < .001. Final Report 50 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 53 shows that a significant correlation (p < .01) exists between teachers’ use of learner response systems and effect size in the continuation study suggesting that as teachers’ utilization of learner response systems increases, effect size tends to increase as well. When combined with the findings from the first year, the correlation was lowered from .49 to .35. The correlation was significant at the .001 level (p < .001). Although reduced slightly, the finding suggests that a moderate positive linear relationship was found between effect size and teachers’ utilization of learner response systems. Evaluation Question 8: What is the relationship between teacher use of the IWB to represent knowledge graphically or nonlinguistically and positive versus negative effects for the use of Promethean ActivClassroom? Representing knowledge graphically or nonlinguistically refers to use of the IWB to present content in ways that do not employ language. High scores on this variable are indicated by use of pictures, pictographs, graphic organizers and the like to present new information. Figure 54 reports the difference in means between the positive-effects and negative-effects classes. Figure 54. Comparison of Means for Use of the IWB for Nonlinguistic Representation Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 3 4 6 1 2 1 1 4.89 (N=18) .40 Negative 1 0 2 0 2 3 0 3.63 (N=8) .65 Positive 19 6 10 14 6 7 13 4.27 (N=75) .25 6 3.21 (N=24) .41 Continuation Combined Negative 2 2 3 2 4 5 t Sig. (2-tailed) df 1.70 .102 24 2.11* .037 97 * p < .05. Figure 54 shows that in the continuation study, the positive-effects group exhibited a higher mean score than the negative-effects group in terms of the utilization of the IWB for nonlinguistic representation. The difference between group means was not significant at the .05 level (p < .05). The combined findings reveal a significant difference (p < .05) between means favoring the positive-effects group. In the continuation study, the mean score for the positive-effects group was above the middle value on the scale. In the positive-effects group, both the median and mode were 5, indicating a central tendency higher than the mean score of 4.89. In the negative-effects group, the median was 3 and the mode was 2. Therefore, the measures of central tendency were at the low end of the scale. In the combined continuation and first year studies, the mean score for the positive-effects group was above the middle value on the scale. The median was 4 and the mode was 7. This indicates a high central tendency in the positive-effects group. In the negative-effects group, the mean was below the middle value on the scale. The median was 3 and the mode was 1. This suggests a low central tendency Final Report 51 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 in the negative-effects group. A reasonable inference can be made that a difference was found between the central tendencies of the observations of utilization of the IWB for nonlinguistic representation which favors the positive-effects group. Figure 55 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 55. Bivariate Correlation for Use of the IWB for Nonlinguistic Representation Pairs Nonlinguistic Representation & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .30 (N=26) .143 Combined .29** (N=99) .004 (2-tailed) ** p < .01. The bivariate correlation for teachers’ use of the IWB for nonlinguistic representation was not significant (p < .05) in the continuation study. When combined with the first year findings, the correlation was significant at the .01 level (p < .01). The combined finding suggests that a moderate positive linear relationship was found between effect size and teachers’ utilization of the IWB for nonlinguistic representation. Evaluation Question 9: What is the relationship between teacher previewing of content and positive versus negative effects for the use of Promethean ActivClassroom? Previewing of content refers to teachers providing students with advanced ways of thinking about a topic before new information is presented. High scores on this variable are manifested by teachers beginning a lesson with activities like: making linkages with what students already know about the topic, providing some type of advance organizer for the content, etc. Figure 56 reports the difference in means between the positive- and negative-effects classes. Figure 56. Comparison of Means for Previewing of Content Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 0 3 3 7 3 2 0 4.11 (N=18) .29 Negative 0 0 1 2 0 1 3 2.57 (N=7) .65 Positive 9 8 5 19 5 15 14 3.61 (N=75) .23 13 2.57 (N=23) .45 Continuation Combined Negative 2 1 2 3 0 2 t Sig. (2-tailed) df 2.51* .019 23 2.16* .034 96 * p < .05. Final Report 52 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 The t-test for the continuation study indicates that the positive-effects group exhibited a higher mean score than the negative-effects group. The difference between group means was statistically significant at the .05 level (p < .05). The combined findings reveal a significant difference (p < .05) between group means favoring the positive-effects group. The mean score for the positive-effects group in the continuation study was close to the middle value on the scale. In the positive-effects group, both the median and mode were 4, indicating a central tendency at the middle of the scale. In the negative-effects group, the median was 2 and the mode was 4. In the negative-effects group, the mean of 2.57 would be the best measure of central tendency. The combined findings show that both group means were below the middle value on the scale. In the positive-effects group, the median and mode were both 4. In the negative-effects group, both the median and mode were 1. This would seem to suggest that there was a medium central tendency in the positiveeffects group and a low central tendency in the negative-effects group. Figure 57 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 57. Bivariate Correlation for Previewing of Content Pairs Previewing of Content & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .63** (N=25) .001 Combined .34** (N=98) .001 (2-tailed) ** p < .01. Figure 57 reveals that a significant positive correlation (p < .01) exists between previewing of content and effect size in the continuation study. This finding would seem to suggest that a strong positive linear relationship was found between effect size and teachers providing students with opportunities to preview content. When combined with the findings from the first year evaluation study, the correlation changed from .63 to .34. The combined correlation was significant at the .01 level (p < .01) indicating that a moderate positive linear relationship was found between effect size and teachers providing opportunities for students to preview content. Evaluation Question 10: What is the relationship between teacher chunking of content and positive versus negative effects for the use of Promethean ActivClassroom? Chunking refers to organizing new content in small segments. High scores on this variable are indicated by teachers presenting content in “digestible bites” and then stopping so that students can process the content. Figure 58 reports the difference in means between the positive- and negative-effects classes for this variable. Final Report 53 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 58. Comparison of Means for Teacher Chunking Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 0 3 5 7 1 3 0 4.21 (N=19) .29 Negative 0 0 0 0 1 7 0 2.13 (N=8) .13 Positive 9 9 19 18 4 14 3 4.30 (N=76) .20 10 1.71 (N=24) .14 Continuation Combined Negative 0 0 0 0 3 11 t Sig. (2-tailed) df 6.57*** .000 23.20 10.80*** .000 91.92 *** p < .001. Levene’s test for equality of variances was significant in both the continuation (F = 7.18, p = .013) and combined studies (F = 15.88, p = .000). As such, Figure 58 reports t-tests for unequal variances. In the continuation study, the t-test reveals a significant difference (p < .001) between mean scores on teacher chunking favoring the positive-effects group. When combined with the first year findings, the difference between means was also significant (p < .001) and favored the positive-effects group. In the positive-effects group in the continuation study, both the median and mode were 4. All measures of central tendency were at the middle of the scale for this variable. In the negative-effects group, the median and mode were 2. All measures of central tendency were at the low end of the scale. In the positive-effects group in the combined study, the median was 4 and the mode was 5. This would suggest a central tendency just above the middle value on the scale. In the negative-effects group, both the median and mode were 2 suggesting a central tendency at the low end of the scale. A reasonable inference can be made that the positive-effects group exhibited more chunking than the negative-effects group. Figure 59 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 59. Bivariate Correlation for Teacher Chunking Pairs Chunking & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .70*** (N=27) .000 Combined .73*** (N=100) .000 (2-tailed) *** p < .001. Figure 59 shows that a significant correlation (p < .001) exists between teacher chunking and effect size in both the continuation and combined studies. These findings suggest that a strong positive linear relationship was found between effect size and teachers’ organization of new content into small segments. Final Report 54 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 11: What is the relationship between teacher scaffolding of content and positive versus negative effects for the use of Promethean ActivClassroom? Scaffolding refers to the extent that “chunks” of information are presented in a way that one leads to another. High scores on this variable are manifest as content being organized in a clear, logical progression. Figure 60 reports the difference in means between the positive-effects and negative-effects classes for this variable. Figure 60. Comparison of Means for Teacher Scaffolding Evaluation Study Frequencies T-Test for Equality of Means Effects Positive Mean SE 7 6 5 4 3 2 1 2 1 6 5 2 3 0 4.32 (N=19) .34 .38 Continuation t Negative 0 0 0 2 1 4 1 2.50 (N=8) Positive 14 7 20 11 8 13 3 4.43 (N=76) .21 8 2.25 (N=24) .27 Combined Negative 0 1 0 4 2 9 Sig. (2-tailed) df 3.11** .005 25 6.40*** .000 52.53 ** p < .01; *** p < .001. The t-test for the continuation study indicates a significant (p < .01) difference between mean scores on teacher scaffolding favoring the positive-effects group. When combined with the first year findings, Levene’s test for equal variances was significant (F = 5.56, p = .020). As a result, the t-test for unequal variances was reported. The combined findings show a significant (p < .001) difference between group means which favors the positive-effects group. In the continuation study, the median in the positive-effects group was 4 and the mode was 5. Therefore, the measures of central tendency are just above the middle value on the scale. In the negative-effects group, the median and mode were 2, indicating that the measures of central tendency were at the low end of the scale. When the findings were combined with the first year evaluation study, both the median and mode were 5 in the positive-effects group suggesting a high central tendency. In the negative-effects group, the median and mode were 2 indicating a low central tendency. A reasonable inference can be made that teachers in the positive-effects group utilized scaffolding more often than teachers in the negative-effects group. Figure 61 reports the bivariate correlation between this variable and the corrected effect sizes. Final Report 55 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 61. Bivariate Correlation for Teacher Scaffolding Pairs Sig. Evaluation Study Correlation Continuation .73*** (N=27) .000 Combined .69*** (N=100) .000 Scaffolding & Corrected Effect Size (2-tailed) *** p < .001. Figure 61 reveals significant (p < .001) positive correlations between scaffolding and effect size in both the continuation and combined studies. These findings imply that a strong positive linear relationship was found between effect size and teachers’ ability to present content that is organized in a clear, logical progression. Evaluation Question 12: What is the relationship between teacher pacing of content and positive versus negative effects for the use of Promethean ActivClassroom? Pacing of content refers to the speed with which the teacher guides students through new information. High scores on this variable are manifest as the teacher slowing down or speeding up a presentation in reaction to students’ needs and engagement levels. Figure 62 reports the difference in means between positive- and negative-effects classes. Figure 62. Comparison of Means for Pacing of Content Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 1 5 4 4 2 4 0 4.35 (N=20) .36 Negative 0 0 0 1 3 2 2 2.38 (N=8) .37 Positive 11 14 17 13 3 17 2 4.45 (N=77) .21 10 2.08 (N=24) .25 Continuation Combined Negative 0 0 1 3 3 7 t Sig. (2-tailed) df 3.21** .004 26 7.38*** .000 57.10 ** p < .01; *** p < .001. Figure 62 shows a significant difference (p < .01) in the continuation study between mean scores on pacing of content that favors the positive-effects group. When combined with the findings from the first year study, Levene’s test was significant (F = 7.89, p = .006). The t-test for unequal variances reveals that the positive-effects group exhibited a higher mean score than the negative-effects group. The difference between means was significant at the .001 level (p < .001). Final Report 56 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 In the positive-effects group in the continuation study, the median was 4.5 and the mode was 6. This would suggest that the central tendency was higher than the mean score of 4.35 for this group. In the negative-effects group, the median was 2.5 and the mode was 3. This indicates that the central tendency was higher than the mean score of 2.38 for this group. In the combined findings, the positive-effects group had a median of 5 and two modes, 2 and 5. The negative-effects group had a median and mode of 2. A reasonable inference can be made that teachers adjusted their pacing more often in the positive-effects group than in the negative-effects group. Figure 63 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 63. Bivariate Correlation for Pacing of Content Pairs Pacing of Content & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .63*** (N=28) .000 Combined .67*** (N=101) .000 (2-tailed) *** p < .001. Figure 63 indicates that positive correlations exist between pacing of content and the corrected effect sizes in both the continuation and combined studies. The bivariate correlations were statistically significant at the .001 level (p < .001). These findings would seem to suggest that a strong positive linear relationship was found between effect size and teachers’ adjustment of the pacing of their content in response to students’ needs and engagement levels. Evaluation Question 13: What is the relationship between teacher monitoring of student progress and positive versus negative effects for the use of Promethean ActivClassroom? Monitoring student progress refers to the teacher systematically identifying students’ level of understanding of the content. High scores on this variable are manifest as the teacher informally or formally assessing students and using the results to alter instruction. Figure 64 reports the difference in means between the positive-effects class and the negative-effects class. Final Report 57 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 64. Comparison of Means for Monitoring Student Progress Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 2 4 5 4 0 5 0 4.45 (N=20) .38 Negative 0 0 0 0 1 4 3 1.75 (N=8) .25 Positive 8 6 20 16 5 16 6 4.01 (N=77) .20 14 1.67 (N=24) .20 Continuation Combined Negative 0 0 0 2 2 6 t Sig. (2-tailed) df 5.93*** .000 25.86 8.32*** .000 72.72 *** p < .001. Levene’s test for equality of means in the continuation study was significant (F = 6.42, p = .018). The ttest for unequal variances indicates that the positive-effects group exhibited a higher mean score on this variable. The difference between means was significant at the .001 level (p < .001). Levene’s test in the combined study was also significant (F = 9.38, p = .003). The t-test for unequal variances shows a significant difference (p < .001) between mean scores which favors the positive-effects group. In the continuation study, there were two modes (2 and 5) in the positive-effects group with a median of 5. The measures of central tendency were higher than the middle value on the scale. In the negativeeffects group, the median and mode were 2. The measures of central tendency were lower than the middle value on the scale. When the findings were combined with the first year study, the mode was 5 in the positive-effects group with a median of 4. In the negative-effects group, both the median and mode were 1. A reasonable inference can be made that monitoring of student progress took place more frequently in the positive-effects group than it did in the negative-effects group. Figure 65 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 65. Bivariate Correlation for Monitoring Student Progress Pairs Monitoring Student Progress & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .79*** (N=28) .000 Combined .72*** (N=101) .000 (2-tailed) *** p < .001. Figure 65 reveals that positive correlations exist between monitoring student progress and the corrected effect sizes in both the continuation and combined studies. The correlations were statistically significant at the .001 level (p < .001). These findings would seem to suggest that a strong positive linear relationship was found between effect size and teachers’ utilization of the results of formal and informal assessments to alter instruction. Final Report 58 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 14: What is the relationship between clarity of content as depicted in the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Clarity of content as depicted on the IWB involves the teacher designing flipcharts so that important content is made clear. High scores on this variable are manifest as the teacher keeping each flipchart free of distracting content and images or the teacher using highlighting features to mark the critical versus peripheral content. Figure 66 reports the difference in means between the positive-effects class and the negative-effects class for this variable. Figure 66. Comparison of Means for Clarity of Content on the IWB Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 3 1 5 4 1 4 0 4.39 (N=18) .41 Negative 0 0 0 1 1 5 1 2.25 (N=8) .31 Positive 13 9 21 12 7 9 4 4.55 (N=75) .20 8 2.21 (N=24) .27 Continuation Combined Negative 0 1 0 4 1 10 t Sig. (2-tailed) df 4.18*** .000 23.24 6.93*** .000 51.69 *** p < .001. In the continuation study, Levene’s test for equality of means was significant (F = 4.33, p = .048). The ttest for unequal variances shows that the positive-effects group had a higher mean score. The difference between group means was significant at the .001 level (p < .001). When combined with the first year findings, Levene’s test was significant (F = 4.44, p = .038). The difference between group means favored the positive-effects group and was statistically significant (p < .001). In the positive-effects group in the continuation study, the median was 4.5 and the mode was 5. The median and mode were higher than the mean of 4.39 suggesting a central tendency higher than the middle value of the scale. In the negative-effects group, the median and mode were 2. All of the measures of central tendency were lower than the middle value of the scale. In the combined findings, the median of the positive-effects group was 5 and the mode was also 5. In the negative-effects group, the median and mode were 2. This seems to suggest that the measures of central tendency were high in the positive-effects group and low in the negative-effects group. A reasonable inference can be made that clarity of content was observed more frequently in the positive-effects group. Figure 67 reports the bivariate correlation between this variable and the corrected effect sizes. Final Report 59 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 67. Bivariate Correlation for Clarity of Content on the IWB Pairs Clarity of Content & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .76*** (N=26) .000 Combined .67*** (N=99) .000 (2-tailed) *** p < .001. In both the continuation and combined studies, the bivariate correlations between clarity of content and effect size were statistically significant at the .001 level (p < .001). These findings would appear to suggest that a strong positive linear relationship was found between effect size and the clarity of content presented with Promethean ActivClassroom. Evaluation Question 15: What is the relationship between student interaction about the content and positive versus negative effects for the use of Promethean ActivClassroom? Student interaction about the content involves students discussing the content with one another. High scores on this variable are indicated when the teacher includes activities like: students answering questions in groups, students summarizing new content in groups, students making predictions in groups, etc. Figure 68 reports the difference in means between the positive- and negative-effects classes. Figure 68. Comparison of Means for Student Interaction Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 0 2 3 4 2 6 3 3.20 (N=20) .37 Negative 0 0 0 0 0 6 2 1.75 (N=8) .16 Positive 7 5 5 15 4 16 25 3.03 (N=77) .23 12 1.96 (N=24) .27 Continuation Combined Negative 0 1 0 3 1 7 t Sig. (2-tailed) df 3.61** .001 24.66 3.00** .004 58.32 ** p < .01. Figure 68 reveals a significant (p < .01) difference between group means favoring the positive-effects group in the continuation study. Levene’s test for equality of means was significant (F = 14.43, p = .001). Therefore, the t-test for unequal variances was reported. When combined with the first year findings, Levene’s test was significant as well (F = 12.11, p = .001). The t-test for unequal variances indicates that the difference between means favored the positive-effects group. The difference was significant at the .01 level (p < .01). Final Report 60 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 In the continuation study, the positive-effects group had a median of 3 and a mode of 2. The median and mode were lower than the mean of 3.20. In the negative-effects group, the median and mode were 2. The median and mode were higher than the mean of 1.75. When combined with the first year findings, the positive-effects group had a median of 2 and a mode of 1. The negative-effects group had a median of 1.5 and a mode of 1. Although the difference between means favored the positive-effects group, all of the measures of central tendency in both groups were lower than the middle value of the scale. Figure 69 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 69. Bivariate Correlation for Student Interaction Pairs Student Interaction & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .57** (N=28) .002 Combined .39*** (N=101) .000 (2-tailed) ** p < .01; *** p < .001. Figure 69 shows a significant correlation (p < .01) between student interaction and effect size in the continuation study. When combined with the findings from the first year study, the correlation changed from .57 to .39. The correlation was statistically significant (p < .001). This finding seems to suggest that a moderate positive linear relationship was found between effect size and student discussion of content. Evaluation Question 16: What is the relationship between student response rates and positive versus negative effects for the use of Promethean ActivClassroom? Student response rates refer to the extent to which students respond to direct queries from the teacher. High scores on this variable manifest as multiple students responding to teacher questions (as opposed to a small group of students) and responses being elicited by the teacher frequently throughout a class period. Figure 70 reports the difference in means between the positive-effects and negative-effects classes for this variable. Final Report 61 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 70. Comparison of Means for Student Response Rate Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 3 1 5 4 3 4 0 4.25 (N=20) .38 Negative 0 0 0 0 1 4 3 1.75 (N=8) .25 Positive 11 7 10 18 11 16 4 4.03 (N=77) .21 14 1.88 (N=24) .28 Continuation Combined Negative 0 1 1 1 2 5 t Sig. (2-tailed) df 5.54*** .000 25.82 5.35*** .000 99 *** p < .001. In the continuation study, Levene’s test was significant (F = 5.80, p = .023). The t-test for unequal variances shows that a significant difference (p < .001) exists between group means favoring the positive-effects group. The combined findings reveal a significant difference (p < .001) between group mean scores which favors the positive-effects group. In the continuation study, the median in the positive-effects group was 4 and the mode was 5. Two of the three measures of central tendency were above the middle value on the scale. In the negative-effects group, the median and mode were 2. All three measures of central tendency were lower than the middle value of the scale. In the combined findings, the median and mode in the positive-effects group were both 4. In the negative-effects group, the median and mode were 1. These findings suggest a central tendency near the middle of the scale in the positive-effects group and a central tendency at the low end of the scale in the negative-effects group. Figure 71 reports the bivariate correlation between this variable and the corrected effect sizes. Figure 71. Bivariate Correlation for Student Response Rate Pairs Student Response Rate & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .73*** (N=28) .000 Combined .67*** (N=101) .000 (2-tailed) *** p < .001. Figure 71 reveals a significant correlation (p < .001) between student response rate and effect size in both the continuation and combined studies. These findings imply that a strong positive linear relationship was found between effect size and student response rate. Final Report 62 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 17: What is the relationship between classroom management and positive versus negative effects for the use of Promethean ActivClassroom? Classroom management refers to the extent to which the teacher uses management strategies that foster a safe and orderly classroom environment. High scores on this variable are manifest as teachers having clear rules and procedures and responding appropriately to classroom interruptions and violations of rules and procedures. Figure 72 reports the difference in means between the positive- and negativeeffects classes. Figure 72. Comparison of Means for Classroom Management Evaluation Study Frequencies T-Test for Equality of Means Effects Mean SE 7 6 5 4 3 2 1 Positive 5 5 7 2 0 0 0 5.68 (N=19) .23 Negative 1 2 4 1 0 0 0 5.38 (N=8) .32 Positive 28 13 21 13 0 0 1 5.68 (N=76) .14 0 5.08 (N=24) .31 Continuation Combined Negative 6 4 5 5 3 1 t Sig. (2-tailed) df .75 .461 25 1.94 .056 98 As shown in Figure 72, the t-test for the continuation study reveals that the positive-effects group exhibited a higher mean score for classroom management than the negative-effects group. The difference between group means was not statistically significant at the .05 level (p < .05). The combined findings also favor the positive-effects group with a difference between means that approaches significance at the .05 level (p < .05). In the positive-effects group in the continuation study, the median was 6 and the mode was 5. The mean score was between the median and mode. In the negative-effects group, the median and mode were 5. The median and mode were below the mean score of 5.38. All measures of central tendency in both groups were higher than the middle value on the scale. When combined with the first year findings, the median in the positive-effects group was 6 and the mode was 7. In the negative-effects group, the median was 5 and the mode was 7. This would seem to suggest that both groups exhibited a high central tendency with respect to teachers having clear rules and procedures and responding appropriately to classroom interruptions and violations of rules and procedures. Figure 73 reports the bivariate correlation between this variable and the corrected effect sizes. Final Report 63 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 73. Bivariate Correlation for Classroom Management Pairs Classroom Management & Corrected Effect Size Sig. Evaluation Study Correlation Continuation .27 (N=27) .169 Combined .30** (N=100) .002 (2-tailed) ** p < .01. The bivariate correlations in the continuation study between classroom management and the corrected effect sizes were not significant at the .05 level (p < .05). The combined findings reveal a significant positive correlation (p < .01). A reasonable inference can be made that a moderate positive linear relationship was found between effect size and classroom management. Interpretation The results from the Phase II analysis provide more detail as to the relationship between the utilization of Promethean ActivClassroom and student achievement. This section focuses on the combined findings from the first year and continuation studies. Four general areas of interest are considered. Comparison of Means versus Bivariate Correlation First, it is worth noting that 15 of the 17 Phase II variables exhibited correlations with the corrected effect sizes from Phase I that were statistically significant at the .05 level (p < .05). It should be further noted that the correlations for all 15 variables would also be considered significant at the .01 level (p < .01). In addition, 12 of the 15 correlations would be considered significant at the .001 level (p < .001). Student IWB skills (Variable 3) and student independent use of the IWB (Variable 5) did not have significant correlations. Of the 15 Phase II variables with significant correlations to effect size, only classroom management (Variable 17) did not exhibit a statistically significant difference on the t-test for equality of means between positive-effects and negative-effects classes (t = 1.94, df = 98, p > .05). However, it should be noted that the t-test approached significance (p = .056) at the .05 level (p < .05). Engagement and Student Achievement Second, the correlation between engagement (Variable 1) and corrected effect size (r = .44, N = 101, p < .001) is consistent with the general finding in Phase I that the use of Promethean ActivClassroom was related to student achievement (see Figure 19). Figure 40 shows that the positive-effects classes had significantly higher mean scores on engagement than the negative-effects classes (t = 3.43, df = 99, p < .01). The measures of central tendency reveal that the positive-effects group had relatively high engagement scores and the negative-effects group had relatively moderate engagement scores (see discussion of Figure 40). Although a significant difference between mean scores on engagement was Final Report 64 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 found, two of the three measures of central tendency in the negative-effects group were above the middle value on the scale suggesting that, in general, students were engaged in both types of classes. Variables 6, 7, and 8 Third, variables 6, 7, and 8 offer insight into the effects of specific features of Promethean ActivClassroom. Variable 6 addressed the extent to which teachers used Promethean ActivClassroom tools to reinforce correct answers produced by students. One interpretation of the correlation between this variable and corrected effect size (r = .40, N = 98, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 16 percentile points on the distribution of corrected effect sizes. Variable 7 addressed the extent to which teachers used learner response systems. One interpretation of the correlation between this variable and corrected effect size (r = .35, N = 101, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 14 percentile points on the distribution of corrected effect sizes. Variable 8 addressed the extent to which teachers used Promethean ActivClassroom to represent knowledge graphically or nonlinguistically. One interpretation of the correlation between this variable and corrected effect size (r = .29, N = 99, p < .01) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 11 percentile points on the distribution of corrected effect sizes. Variables 10, 11, 12, 13, 14, and 16 Fourth, an examination of the effects of variables 10, 11, 12, 13, 14, and 16 provides insight into general instructional variables that might interact with use of Promethean ActivClassroom. All of these variables exhibited correlations with effect size that were greater than .60 and could be considered large effects in the social sciences (see Lipsey & Wilson, 2001, p. 147). Variable 10 addressed chunking of content. One interpretation of the correlation between this variable and corrected effect size (r = .73, N = 100, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 27 percentile points on the distribution of corrected effect sizes. Variable 11 addressed scaffolding. One interpretation of the correlation between this variable and corrected effect size (r = .69, N = 100, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 26 percentile points on the distribution of corrected effect sizes. Final Report 65 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Variable 12 addressed pacing. One interpretation of the correlation between this variable and corrected effect size (r = .67, N = 101, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 25 percentile points on the distribution of corrected effect sizes. Variable 13 addressed monitoring of student progress. One interpretation of the correlation between this variable and corrected effect size (r = .72, N = 101, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 26 percentile points on the distribution of corrected effect sizes. Variable 14 addressed the clarity of the content as depicted on the IWB. One interpretation of the correlation between this variable and corrected effect size (r = .67, N = 99, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 25 percentile points on the distribution of corrected effect sizes. Variable 16 addressed student response rates. One interpretation of the correlation between this variable and corrected effect size (r = .67, N = 101, p < .001) is that if a typical teacher’s use of this tool were to increase by one standard deviation, the corrected effect size would be predicted to increase by 25 percentile points on the distribution of corrected effect sizes. Figure 74 displays the product moment correlations between these variables. Final Report 66 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Figure 74. Product Moment Correlations between Variables 10, 11, 12, 13, 14, and 16 Comparison Correlation Variable 10 Variable 11 Variable 12 Variable 13 Variable 14 Variable 16 Variable 10 Variable 11 Variable 12 Variable 13 Variable 14 Variable 16 1 .86*** .81*** .78*** .81*** .72*** .000 .000 .000 .000 .000 Sig. (2-tailed) N 100 100 100 100 99 100 Correlation .86*** 1 .78*** .74*** .82*** .65*** Sig. (2-tailed) .000 .000 .000 .000 .000 N 100 100 100 100 99 100 Correlation .81*** .78*** 1 .71*** .76*** .66*** Sig. (2-tailed) .000 .000 .000 .000 .000 N 100 100 101 101 99 101 Correlation .78*** .74*** .71*** 1 .66*** .70*** Sig. (2-tailed) .000 .000 .000 .000 .000 N 100 100 101 101 99 101 Correlation .81*** .82*** .76*** .66*** 1 .61*** Sig. (2-tailed) .000 .000 .000 .000 N 99 99 99 99 99 99 Correlation .72*** .65*** .66*** .70*** .61*** 1 Sig. (2-tailed) .000 .000 .000 .000 .000 N 100 100 101 101 99 .000 101 *** p < .001. Figure 74 indicates that these variables are highly intercorrelated. Stated differently, the high correlations imply that these instructional practices are somehow linked in terms of their implementation in the classroom. In order to assess their composite effect, a macro written for meta-regression in SPSS was used to conduct a modified weighted least squares regression analysis (Wilson, 2005). This approach was taken because the weighting procedure employed by SPSS makes an assumption that the weights represent different subjects as opposed to studies, and as such, statistical significance tests are based on sample size assumptions that would not apply to meta-analytic data (see Lipsey & Wilson, 2001). A mixedeffects model was utilized in the meta-regression analysis with a maximum likelihood estimate of the random-effects variance component. Final Report 67 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 The corrected effect sizes from Phase I were the dependent variable in the meta-regression and the ratings on chunking (Variable 10), scaffolding (Variable 11), pacing (Variable 12), monitoring (Variable 13), clarity of IWB (Variable 14), and student response rate (Variable 16) were the independent variables. The inverse of the variance (1/SE2) was used to weight the corrected effect sizes. The results of this analysis are depicted in Figure 75. Figure 75. Weighted Meta-Regression Analysis with Corrected Effect Size as the Dependent Variable and Variables 10, 11, 12, 13, 14, and 16 as Independent Variables R R2 Adj. R2 N .789 .622 .597 99 Q df Sig. (2-tailed) Model 158.745**** 6 .0000 Residual 96.563 92 .3520 Total 255.308**** 98 .0000 Model Summary v = .06; se(v) = .02. **** p < .0001. Figure 75 shows that the overall model was statistically significant at the .0001 level (p < .0001). Applying a widely used convention for appraising the magnitude of effect sizes, the multiple correlation between the observed values of the dependent variable (i.e., corrected effect size) and the predicted values (R = .789, N = 99, p < .0001) would be considered very large in the social sciences (see Lipsey & Wilson, 2001, p. 147). Approximately 62% of the variability in the dependent variable was accounted for by the variables in this model (R2 = .622). After taking into account the number of predictors, about 60% of the variability was accounted for by the variables in the model (Adj. R2 = .597) with about 40% of the variance attributable to other factors not examined in this report. Phase II Summary and Interpretation Phase II of the evaluation study regarding the effects of Promethean ActivClassroom on student achievement sought to answer the following questions: Evaluation Question 1: What is the relationship between student engagement and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 2: What is the relationship between teacher IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 3: What is the relationship between student IWB skills and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 4: What is the relationship between multiple student use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Final Report 68 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Evaluation Question 5: What is the relationship between student independent use of the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 6: What is the relationship between teacher use of IWB reinforcers and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 7: What is the relationship between teacher use of learner response systems and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 8: What is the relationship between teacher use of the IWB to represent knowledge graphically or nonlinguistically and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 9: What is the relationship between teacher previewing of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 10: What is the relationship between teacher chunking of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 11: What is the relationship between teacher scaffolding of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 12: What is the relationship between teacher pacing of content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 13: What is the relationship between teacher monitoring of student progress and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 14: What is the relationship between clarity of content as depicted in the IWB and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 15: What is the relationship between student interaction about the content and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 16: What is the relationship between student response rates and positive versus negative effects for the use of Promethean ActivClassroom? Evaluation Question 17: What is the relationship between classroom management and positive versus negative effects for the use of Promethean ActivClassroom? Phase II involved an analysis of student learning with and without Promethean ActivClassroom from the perspective of teacher behavior as evidenced in videotapes of teachers using the technology in their classrooms. The general focus of Phase II was to determine the behaviors that differentiated those teachers who obtained positive effects from Promethean ActivClassroom from those who did not. The combined findings from the first year and continuation studies can be summarized in the following statements: • • The correlations between 15 of the 17 variables addressed in Phase II and the corrected effect sizes calculated in Phase I were statistically significant at the .05 level (p < .05). The correlations for all 15 variables would also be considered significant at the .01 level (p < .01). In addition, 12 of the 15 correlations would be considered significant at the .001 level (p < .001). Student IWB skills (Variable 3) and student independent use of the IWB (Variable 5) did not have significant correlations. Of the 15 Phase II variables with significant correlations to effect size, only classroom management (Variable 17) did not exhibit statistical significance on the t-test for equality of means at the .05 level (p < .05). However, the t-test did approach significance (t = 1.94, df = .98, p = .056). Final Report 69 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 • • The correlations between 14 of the 17 Phase II variables and the corrected effect sizes were greater than .30. Of those 14 variables, 6 exhibited correlations with corrected effect size that were greater than .60, which would seem to suggest that substantial increases in student achievement would be predicted with improvements in teacher behavior with respect to chunking (Variable 10), scaffolding (Variable 11), pacing (Variable 12), monitoring (Variable 13), clarity of IWB (Variable 14), and student response rate (Variable 16). The multiple correlation of .789 (N = 99, p < .0001) reported in Figure 75 might suggest a strong effect on student achievement when the following conditions are met: o New content is organized into small digestible bites designed with students’ background knowledge and needs in mind (chunking). o Chunks of new content follow a logical progression so that each chunk helps students understand the next (scaffolding). o The pace at which each chunk is addressed is adjusted as needed (i.e., slower, faster) to maintain high engagement and comprehension (pacing). o Students’ ability to understand new content is consistently monitored (monitoring). o Promethean ActivClassroom is utilized so that essential content is presented in a clear manner. Content should be reviewed or re-taught when it becomes apparent that students misunderstand portions of the content (clarity of IWB). o Questions are asked and addressed in a manner that would allow all students to have an opportunity to respond. Students’ answers should be continually examined for correctness and depth of understanding (student response rate). Final Report 70 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Technical Notes Technical Note 1: Conceptually, analysis of covariance (ANCOVA) can be thought of as using a covariate to predict performance on the dependent variable and then using the residual scores (predicted score minus observed score) as the dependent measure in the analysis. For this report, students’ posttest scores were predicted using the scores received on the pretest. The difference between the predicted posttest scores and the observed posttest scores were then computed for each student that took both pretest and posttest. This difference is referred to as the residual score for each student. It represents the part of each student’s posttest score that cannot be predicted from the pretest score for that student. Theoretically, use of residual scores based on pretest predictions is an attempt to equate all students on the dependent measure prior to execution of the intervention—in this case the use of Promethean ActivClassroom. Technical Note 2: The quasi-experimental design used in the treatment/control studies is referred to as a pretest-posttest non-equivalent groups design. The groups are considered to be non-equivalent because students were not assigned to treatment and control conditions through random lottery. Pretest scores were used as a covariate to statistically equate the students and partially control for differing levels of background knowledge and skill. Although ANCOVA was used to statistically equate students in terms of prior achievement, without randomization arguments about causal relationships are severely weakened. While quasi-experimental designs lack the rigor of fully randomized experiments, the pretest-posttest nonrandomized research design has a natural fit with classroom teachers. This design allows teachers to use one intact class as a control group and a second intact class as the treatment group. Potential concerns over making adjustments to existing lesson plans or instructional time in order to isolate control group students from treatment group students in a single class would not apply. In other words, the use of this quasi-experimental design minimizes the potential impact on a teacher’s normal class routine. An assumption was made that both groups in the treatment/control studies shared a similar demographic (e.g., ethnicity, socio-economic status, etc.) which would allow for a fair comparison. This notwithstanding, random assignment of students to treatment and control groups is always preferable in evaluation studies. This effort does not meet that criterion. Technical Note 3: The generic term effect size applies to a variety of indices (e.g., d, r, R, PV) that can be used to demonstrate the effect of an independent variable (e.g., use of Promethean ActiVote) on a dependent variable (e.g., student academic achievement). As used in this report, effect size refers to the standardized mean difference effect size. This index, first popularized by Glass (1976) and Cohen (1977), is the difference between experimental and control group means divided by an estimate of the population standard deviation. mean of treatment group – mean of control group standardized mean difference effect size = estimate of population standard deviation Final Report 71 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Consider the following illustration of the use of effect size. Assume that the achievement mean of a group of students in a class that used a specific instructional strategy (e.g., graphic organizers) is 90 on a standardized test and the mean of a group of students in a class that did not use the instructional strategy is 80. Assuming the population standard deviation is 10, the effect size would be as follows: 𝐸𝑆 = 90 − 80 = 1.0 10 Thus, an effect size of 1.0 would indicate that the mean score in the treatment group is one standard deviation higher than the mean score in the control group. Conversely, an effect size of -1.0 would indicate that the mean score in the treatment group is one standard deviation lower than the mean score in the control group. Technical Note 4: The meta-analytic findings in this report are typically reported in two ways— observed and corrected. The corrected findings have been corrected for attenuation due to a lack of reliability in the dependent measure (i.e., teacher-designed assessments of student academic achievement). Hunter and Schmidt detail the rationale and importance of correcting for 11 attenuation artifacts—one of which is random error associated with measurement of the dependent variable (2004, pp. 301-313). They explain: . . . error of measurement in the dependent variable reduces the effect size estimate. If the reliability of measurement is low, the reduction can be quite sizable. Failure to correct for the attenuation due to error of measurement yields an erroneous effect size estimate. Furthermore, because the error is systematic, a bare-bones meta-analysis on uncorrected effect sizes will produce an incorrect estimate of the true effect size. The extent of the reduction in the mean effect size is determined by the mean level of reliability across the studies. Variation in reliability across studies causes variation in the observed effect size above and beyond that produced by sampling error. . . . A bare-bones meta-analysis will not correct for either the systematic reduction in the mean effect size or the systematic increase in the variance of effect sizes. Thus, even meta-analysis will produce correct values for the distribution of effect sizes only if there is a correction for the attenuation due to error of measurement. (p. 302) For ease of discussion we consider correcting for attenuation due to unreliability in the dependent measure using the population correlation instead of the population standardized mean difference effect size. The reader should note that the example provided regarding correcting correlations is analogous to correcting a standardized mean difference. To illustrate correcting for attenuation due to unreliability in the dependent measure, assume that the population correlation between a specific instructional strategy (e.g., nonlinguistic representations) and student academic achievement is .50. A given study attempts to estimate that correlation but employs a measure of the dependent variable (i.e., a teacher-designed assessment of student academic achievement) that has a reliability of .81—considered a typical reliability for a test of general cognitive ability. According to attenuation theory, the correlation would be reduced by the square root of the reliability (i.e., the attenuation factor). In other words, the population correlation is multiplied by the attenuation factor (√. 81 = .90), thus reducing the correlation by 10 percent. Therefore, the observed correlation will be .45 (.50 x .90) even if there is no attenuation due to the other ten artifacts listed by Hunter and Schmidt (2004, p. 35). When the measure of the Final Report 72 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 dependent variable has a lower reliability, .36 for example, the correlation is reduced by 40 percent (√. 36 = .60) to .30 (.50 x .60). In order to make a correction for attenuation, the correlation is divided by the attenuation factor (i.e., the square root of the reliability). (For a more detailed discussion of attenuation in the context of meta-analysis see Hunter and Schmidt, 2004.) For the purposes of this report, an estimate of reliability was used. Osborne, Christensen, and Gunter reported that the average reliability reported in top Educational Psychology journals was .83 (as cited by Osborne, 2003). Lou and colleagues (1996) reported a typical reliability of .85 for standardized achievement tests and a reliability of .75 for unstandardized achievement tests. Because the dependent measure in the independent action research studies involved teacher-designed assessments of student academic achievement, .75 was used as the reliability to correct for attenuation using the following formula: 𝑑𝑐 = 𝑑𝑜 𝑎 In the formula, 𝑑𝑐 is the corrected effect size, 𝑑𝑜 is the observed effect size, and 𝑎 is the attenuation factor. Using this formula, each effect size reported in Figure 18 was corrected for attenuation to produce the corrected meta-analytic findings considered in this report. Technical Note 5: Independent variables can be analyzed as fixed effects or as random effects. In the context of ANOVA/ANCOVA, fixed effects are factors with levels that are deliberately arranged by the researcher. In contrast, random effects are factors with levels that are not deliberately arranged. Instead, random effects are factors which are randomly sampled from a population of possible samples. Generally speaking, when independent variables are analyzed as random effects, the intent is to generalize results beyond the boundaries of the independent variables employed in the study. For example, if a researcher were interested in the effect that the quality of school leadership has on academic proficiency, the researcher could select a random sample of schools in order to estimate the amount of variance in student academic achievement attributable to differences between types of school leaders. Thus, using the sample, the researcher can make generalizations regarding the influence of school leadership on academic achievement as a whole. Additional research could attempt to replicate the findings by duplicating the study and selecting a different random sample of schools for comparison. If the findings could be replicated then the generalization would have increased validity. However, if the findings could not be replicated then the generalization might not be valid across all settings. An alternative hypothesis might need to be explored. When fixed effects are employed one typically does not generalize beyond the boundaries of the independent variables in the study. However, additional research could still attempt to generalize the findings by replicating every facet of the study. For example, multiple studies could be used to determine the influence of a specific instructional technique in the classroom. Final Report 73 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Technical Note 6: In Figure 18, the column labeled “ES” contains the computed effect size for each study calculated as Cohen’s d using the following formula: 𝑑= 𝑟 �(1 − 𝑟 2 )(𝑝(1 − 𝑝)) In the formula r is the effect size correlation (e.g., point-biserial correlation coefficient) and p is the proportion of the total population in one of the two groups (i.e., the treatment group). This formula is used to compute the standardized mean difference effect size from an effect size correlation when the treatment and control group populations are not equal (see Lipsey & Wilson, 2001, pp. 62-63). ANCOVA was employed for each treatment/control study. Partial eta squared (η2𝑝 ) is a measure of strength of association between independent variables generated by SPSS for the ANCOVA (SPSS, 2009). Another measure of strength of association, eta squared (η2 ), is commonly used to describe the proportion of total variability in a dependent variable (e.g., student’s posttest scores) that can be explained by an independent variable (e.g., the use of Promethean ActivClassroom). Whereas, eta squared (η2 ) describes the proportion of the total variance attributable to an effect, partial Eta squared (η2𝑝 ) describes the proportion of the effect plus error that is attributable to the effect. These two measures are often depicted by the following formulas: 𝜂2 = 𝑆𝑆𝑒𝑓𝑓𝑒𝑐𝑡 2 𝑆𝑆𝑒𝑓𝑓𝑒𝑐𝑡 , 𝜂𝑝 = 𝑆𝑆𝑡𝑜𝑡𝑎𝑙 𝑆𝑆𝑒𝑓𝑓𝑒𝑐𝑡 + 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 In the formulas, SSeffect is the sum of squares for whatever effect is of interest, SSerror is the sum of squares for whatever error term is associated with that effect, and SStotal is the total sums of squares for all effects of interest, interactions, and errors in the ANCOVA. When there is one degree of freedom, the square root of eta squared (η2 ) is r. Because only one independent variable was considered in the ANCOVA, the square root of partial eta squared (η2𝑝 ) was used as a reasonable estimate for r in the formula for computing d above. Technical Note 7: The standardized mean difference effect size (Cohen’s d) expresses the differences between means in standardized or “Z score” form, which gives rise to another index frequently used in research regarding education—percentile gain. By way of illustration, consider the example in Technical Note 3. An effect size (Cohen’s d) of 1.0 can be interpreted as the average score in the treatment group being about 34 percentile points greater than the average score in the control group. It is important to note that this does not mean that the average score in the treatment group is 34% greater than the average score in the control group. Instead, it refers to the percentile point on the distribution of scores. For example, a score at the 65th percentile would be the lowest score that is greater than 65% of the scores in the distribution. Again, the effect size translates the difference between group means into Z score form. Distribution theory dictates that a Z score of 1.0 is at the 84.13 percentile point of the standard normal distribution. To determine the percentile gain, the effect size is transformed into percentile points above or below the 50th percentile point on the unit Final Report 74 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 normal distribution (e.g., 84.1 - 50.0 = 34.1). The percentile gain values listed in this report were determined by consulting a normal curve table for the area for each reported effect size. Using the current example of a study with an effect size (Cohen’s d) of 1.0, the average student in the treatment group would be expected to be at the 84th percentile in the control group, which is 34 percentile points greater than the average student in the control group. Stated differently, the average student in the treatment group (i.e., the student at the 50th percentile) scored better than 84% of the students in the control group. An alternate interpretation involves a student who is ranked in the middle of a control group with 100 students. If this student were the only one to receive the treatment, his or her ranking would improve from 50th to 17th. In other words, the student would be expected to surpass 33% of the students that did not receive the treatment. Sometimes it can be helpful to visualize the effect size in terms of the overlapping distributions of scores from the two groups. One of the most widely used approaches is the binomial effect size display (BESD). This approach assumes a success threshold at the median of the distribution of scores on a dependent variable for the entire population (i.e., combined treatment and control groups). This threshold is always set at 50% for the combined distributions. The BESD can be used to show the proportion of students who would be expected to pass or fail a hypothetical test represented by the dependent variable assuming an initial passing rate of .50 for all students. In order to utilize the BESD with the standardized mean difference effect size (Cohen’s d) it must first be transformed into its correlational equivalent using the following formula: 𝑟= 𝑑 √4 + 𝑑 2 The correlation is divided by 2 and added to or subtracted from .50 (i.e., .50 ± r /2). Figure TN 7.1 displays the BESD for an effect size (Cohen’s d) of 1.0. Figure TN 7.1. Binomial Effect Size Display (BESD) Group Expected to Pass Expected to Fail Treatment .725 .275 Control .275 .725 Taken at face value, the BESD for an effect size (Cohen’s d) of 1.0 would suggest that 72.5% of the students in the treatment group would be expected to pass the hypothetical test. In comparison, only 27.5% of the students in the control group would be expected to pass. In other words, the BESD indicates a success rate of 27.5% for the control group and 72.5% for the treatment group. It is important to note that the BESD assumes a 50% base rate for both groups, creating an artificial situation to illustrate the impact of the effect. That said, it does allow an effect size to be visualized in what some may consider a more intuitive binomial display. Final Report 75 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 A more conservative variation on the BESD format sets the success threshold at the median of the lower distribution. Whenever an effect size is positive, the success rate is set to 50% for the control group. If the effect size were negative, the success rate would be set to 50% for the treatment group. Figure TN 7.2 depicts this variation. Figure TN 7.2. A Variation of the Binomial Effect Size Display (BESD) Group Above Lower Group Median Below Lower Group Median Treatment .841 .159 Control .500 .500 Figure TN 7.2 shows a success rate of 84.1% in the treatment group compared with a success rate of 50.0% in the control group. In other words, 84.1% of the students in the treatment group scored higher on the hypothetical test than the control group median. In contrast, 50.0% of the students in the control group scored higher than this value which represents a gain of 34.1 percentile points in the treatment group. Again, using this approach the base rate is always set at 50% for the lower group. This approach parallels the description of percentile gain above. However, instead of representing gain (or loss) in terms of the treatment group median (i.e., the student at the 50th percentile), this approach depicts the difference between groups in terms of the median of the group with the lower distribution. The percentile gains presented in this report can be thought of either way. For example, an effect size (Cohen’s d) of 1.0 can be interpreted to mean that the student at the 50th percentile in the treatment group scored higher than 84.1% of the students in the control group, or it can mean that 84.1% of the students in the treatment group scored higher than the student at the 50th percentile in the control group. (For a more thorough discussion regarding interpretation of effect size, see Lipsey & Wilson, 2001; Cooper, 2009.) Technical Note 8: Within the context of meta-analysis, independent studies can be analyzed using a fixed-effect or random-effects model. It is important to note that in a meta-analysis the terms fixed and random do not have the same meaning as they do with respect to independent variables in primary studies (see Technical Note 5). Fixed-effect models are based on an assumption of one true treatment effect common to every study. In other words, fixed-effect models assume that exactly the same effect size underlies all studies in the meta-analysis. Additionally, fixed-effect models are based on an assumption that differences in observed effects are due to sampling error alone. Random-effects models do not assume the existence of a common treatment effect. In contrast, random-effects models allow for the possibility that the effect size varies from study to study. Stated differently, random-effects models are based on an assumption that the true treatment effect sizes in the individual studies may differ from one another. Random-effects models are often used to estimate this variance. (For a more thorough discussion regarding models used in meta-analysis, see Lipsey & Wilson, 2001; Hunter & Schmidt, 2004; Cooper, 2009; Borenstein, Hedges, Higgins, & Rothstein, 2009.) Final Report 76 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Technical Note 9: In a meta-analysis, effect sizes from multiple studies on a common topic are averaged together to determine the effect for a specific treatment (e.g., use of Promethean ActivClassroom). Metaanalysts are interested in determining why effect sizes vary from one comparison to another. In order to accomplish this, the meta-analyst performs a homogeneity analysis. A homogeneity analysis compares the observed variance between effect sizes with the variance that would be expected from sampling error alone. It provides a mechanism for determining the probability of the observed variance being attributed to sampling error. A homogeneous distribution of effect sizes would indicate that individual effect sizes differ from the population mean solely due to sampling error. For some meta-analysts there would be little value in examining study characteristics to explain observed variation in effect sizes. After all, sampling error would be the simplest explanation for variation in effects. A heterogeneous distribution would suggest that there are differences among effect sizes that have a source of difference (e.g., differences associated with study characteristics) other than sampling error. Meta-analysts would then examine whether study characteristics are systematically associated with variation in effects. In order to test whether a set of effect sizes is homogeneous, meta-analysts calculate a statistic Hedges and Olkin (1985) called Qt. The statistic Qt is distributed as a chi-square with one less than the number of effect sizes as degrees of freedom. If Qt is statistically significant at the chosen level (e.g., p < .05) the effect sizes are heterogeneous and the meta-analysts reject the null hypothesis that sampling error alone produced the variance in effect sizes. A reasonable inference can be made that there is another source for the difference. If Qt is not statistically significant, the effect sizes are homogeneous and the null hypothesis that sampling error alone produced the variance in effect sizes cannot be rejected. Therefore, sampling error alone is the most parsimonious explanation for the difference in effects. In order to test whether a set of average effect sizes is homogeneous, meta-analysts calculate a statistic called Qb. The statistic Qb is used to test whether the average effects from specific groupings (i.e., on a specified moderator variable) are homogeneous. Qb is distributed as a chi-square with one less than the number of groupings as degrees of freedom. If Qb is statistically significant at the chosen level (e.g., p < .05) the average effect sizes are heterogeneous and the meta-analysts reject the null hypothesis that the effect sizes were equal in the populations for each group. In other words, the grouping variable can be considered a significant contributor to the variance in effect sizes. If Qb is not statistically significant, the average effect sizes are homogeneous and the null hypothesis that the effect sizes were equal in the populations for each group cannot be rejected. The grouping variable would not explain the variance in effects beyond that associated with sampling error. (For a detailed discussion of homogeneity analysis and the Q-statistic, see Lipsey & Wilson, 2001; Cooper, 2009; Borenstein, Hedges, Higgins & Rothstein, 2009.) Technical Note 10: The independent samples t-test compares the mean scores between two groups on a given variable. The t-test is based on three assumptions: 1) the dependent variable is normally distributed, 2) the two groups have approximately equal variance on the dependent variable, and 3) the two groups are independent of each other. If the t-test is found to be statistically significant at the standard alpha of .05 (p < .05) the null hypothesis that there is no difference between means can be rejected. In other words, the two group means would be considered statistically different from each Final Report 77 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 other. In order to avoid a violation of the second assumption, Levene’s test is used to determine if the two groups have equal variance. If Levene’s test is found to be statistically significant at an alpha of .05 (p < .05) then the SPSS output for the t-test for unequal variances is utilized. The t-test for unequal variances is based on a different formula and a different calculation for degrees of freedom. Technical Note 11: In order to correct for attenuation each of the computed effect sizes in Phase I were divided by the same attenuation factor (see Technical Note 4). Therefore, the bivariate correlations between the Phase II variables and the observed and corrected effect sizes were identical. Consequently, only the correlations with the corrected effect sizes were reported. Final Report 78 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Appendix A – MRL Action Research Instructions Promethean ActivClassroom Evaluation Study As an MRL action research participant, you will be asked to select a specific unit of instruction, or set of related lessons on a single topic (hereinafter referred to as unit) to be administered to two different groups of students—an experimental group and a control group. At the beginning of the unit, you would design and administer the same comprehensive pretest to both experimental and control groups over the content that will be covered. If this is not a common practice, you might want to explain to students that they are not expected to do well on the pretest since the material has not yet been covered; it is simply a test to find out what they know prior to instruction so that it might be determined how much they have learned as a result of instruction. Regardless of how many items used on the pretest or the weights assigned those items, when scoring the tests, each student’s score should be converted to a percentage. At the end of the unit the same comprehensive posttest would be given to both groups of students. If the pretests are not returned to students, the same comprehensive assessment used at the beginning of the unit may be used again. If the pretests are returned to students, there should be changes made in the posttest to ensure that students can’t simply recall the answers from the pretest. Like the pretest, the posttest should be scored using a percentage scale. Again, you are teaching the same unit to two different groups of students. Ideally, you would teach the unit to both groups concurrently ensuring that the duration of the unit is the same for both groups. One class period would serve as the experimental group and a second class period would serve as the control group. When teaching the unit to the experimental group, you would make sure you use Promethean ActivClassroom whenever and in ways you believe it to be applicable. When teaching the unit to the control group, you would not use Promethean ActivClassroom. If you are an elementary school teacher and do not have two different classes of students then you would teach two consecutive units to the same group of students. The first unit is considered the experimental group in that Promethean ActivClassroom is used. The second unit is considered the control group in that Promethean ActivClassroom is not used. Please ensure that the duration of each unit is the same and that the content of each unit is as closely related as possible without being identical or duplicated (e.g., complementary topics in mathematics). If you choose to team up with a second teacher (one teacher serving as the experimental group and the other teacher serving as the control) your unit must be designed in such a way that all facets of instruction are as close to the same as possible (e.g., the same lesson plan, duration of lesson, class materials, tests administered, etc.). Forms submitted to MRL need to indicate this approach so that it may be reflected in the written report. When the posttest is completed, each student’s pretest and posttest percentage scores should be recorded on Form A along with general demographic information for each student. If a student does not take a test, leave a blank space on the form to indicate a missing test. Note that the form does not ask for each Final Report 79 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 student’s name. This has been done intentionally to comply with student privacy requirements. In short, make sure that each row of Form A contains a specific student’s pretest percentage score (without the percentage sign, i.e., 90% = 90), the same student’s posttest percentage score (without the percentage sign), and the student’s general demographic information. Please use the ethnicity codes listed at the bottom of each form when filling out the demographic information for your students. Again, do not record student names or ID codes on the form. For the purposes of this action research project, please refer to the following definitions when filling out the forms: • • • Free/Reduced Lunch (FRL): students who receive free or reduced-price lunches through the National School Lunch Program (NSLP). English Language Learner (ELL): students who have been designated as Limited English Proficient (LEP), English as a Second Language (ESL), or Second Language Learner (SLL). Students that have been reclassified as fluent English speakers should not be included in this demographic category. Special Education (Sp. Ed.): students who receive special education and related services under the Individuals with Disabilities Education Act (IDEA). At the end of the experimental/control units, you will be asked to collect data from students regarding their perceptions of the unit. To do this, you should present students in both groups with the appropriate student questionnaire (Form B—lower elementary, upper elementary, middle school, high school). This can be done right before or after students take the posttest. In any event, this form must be completed prior to students receiving their scores on the posttest so that those scores do not influence them when they complete the form. You should go over the directions with students, emphasizing that responses will not be able to be tracked back to individual students. To this end, it is best to have students complete the form and then immediately place all forms in a large envelope that is sealed in front of them. The next type of data to be collected comes from you, the classroom teacher. Form C asks you to respond to seven questions regarding your experience as a teacher and your use of Promethean ActivClassroom. Finally, Form D asks for general information regarding your unit of instruction— topic, duration, etc. If you chose to team up with another teacher, both teachers need to submit Forms C and D. Please do not put both teacher ID codes on Form A. Submit them as one action research project with the ID code for the teacher that taught each group on the appropriate scoring sheet. It is important to note that this data will be collected and reported in a manner that provides anonymity for teachers and students. You will be asked to provide a personal alpha-numeric ID code as opposed to your name. Only you will know which results apply to your students. This is important, as supplying your name removes the anonymity from the action research project. Do not use your name, or any part of your name, as your personal ID code. As an alternative your project leader can randomly assign an ID code based on the school or district name and a number. For example, two teachers from Some City High School could be assigned SomeCity1 and SomeCity2 or SCHS1 and SCHS2. To maintain anonymity, the ID codes should be determined at the start of the action research project and given to each teacher at random to ensure that only the teachers involved know their respective ID codes. Electronic versions of Forms A, C, and D will be given to your project leader. Please fill out the electronic versions of these forms and print a hard copy. Doing so will help minimize data entry errors Final Report 80 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 and allow for efficient import into statistical software for data analysis. When all forms have been filled out (i.e., Form A with students’ pretest and posttest scores and demographics for experimental and control groups, copies of Form B from each student, and Forms C and D filled out by you) put the hard copies in a large envelope and turn the envelope into your project leader. Your project leader will determine how the electronic forms will be collected (e.g., email attachment, CDROM, etc.). After data has been collected and submitted to MRL, it will be analyzed and a report will be generated similar to those found on MRL’s website (see http://www.marzanoresearch.com/ research/actionresearch_reports.aspx). The report will present data for each teacher listed by an assigned numeric code. Your project leader will be provided with a list matching the assigned codes to the chosen ID codes for your school. In this way, you might anonymously compare the pretest/posttest gains of your students with those of other participating teachers at your school and across the evaluation study. This data can be used to stimulate discussion regarding how to maintain expertise as professional educators and better serve the needs of students. In addition to these requirements, you need to videotape a single lesson (of your choosing) from the unit for both experimental and control groups. For accurate comparison, the videotaped lesson must be the same for both groups. The general purpose of the videotapes will be to determine the behaviors that differentiate those teachers who obtain positive effects from Promethean ActivClassroom from those who do not. The video camera should be set up in such a way as to capture you and your students’ interaction with the Promethean technology, perhaps off to one side of the classroom. It would be helpful, if someone could periodically pan the classroom, to capture as much of the class as possible. Please follow established guidelines for videotaping students set forth by your school. A sample parental consent form has been included in this packet should you require one. If collected from parents do not send them to MRL. MRL can analyze a variety of common digital camcorder media such as MiniDV tape and Mini DVD. If feasible, it would be helpful if each teacher’s videotapes were combined and transferred to a DVD-R disc (using each teacher’s personal ID code to identify his or her respective video content) and then included in the data package sent by your project leader to MRL. Thank you again for considering involvement in an action research project. Final Report 81 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Appendix B – NCES Code Definitions Figure B.1. NCES Codes and Definitions NCES Locale Code & Category Definitions Used by NCES 11 – City, Large Territory inside an urbanized area and inside a principal city with population of 250,000 or more. 12 – City, Midsize Territory inside an urbanized area and inside a principal city with population less than 250,000 and greater than or equal to 100,000. 13 – City, Small Territory inside an urbanized area and inside a principal city with population less than 100,000. 21 – Suburb, Large Territory outside a principal city and inside an urbanized area with population of 250,000 or more. 22 – Suburb, Midsize Territory outside a principal city and inside an urbanized area with population less than 250,000 and greater than or equal to 100,000. 23 – Suburb, Small Territory outside a principal city and inside an urbanized area with population less than 100,000. 31 – Town, Fringe Territory inside an urban cluster that is less than or equal to 10 miles from an urbanized area. 32 – Town, Distant Territory inside an urban cluster that is more than 10 miles and less than or equal to 35 miles from an urbanized area. 33 – Town, Remote Territory inside an urban cluster that is more than 35 miles from an urbanized area. 41 – Rural, Fringe Census-defined rural territory that is less than or equal to 5 miles from an urbanized area, as well as rural territory that is less than or equal to 2.5 miles from an urban cluster. 42 – Rural, Distant Census-defined rural territory that is more than 5 miles but less than or equal to 25 miles from an urbanized area, as well as rural territory that is more than 2.5 miles but less than or equal to 10 miles from an urban cluster. 43 – Rural, Remote Census-defined rural territory that is more than 25 miles from an urbanized area and is also more than 10 miles from an urban cluster. Source: From NCES Common Core of Data (http://nces.ed.gov/ccd/rural_locales.asp). Final Report 82 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Appendix C – Treatment/Control Study Data Figure C.1. Year Two Continuation Study Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 1 -.64 1 Elementary School Social Studies -- -- -- -- 2 -.30 4 Elementary School Language Arts 28.00 36.00 35 5 3 1.03 3 Elementary School Language Arts 3.00 6.00 75 4 4 -1.24 4 Elementary School Language Arts 15.00 36.00 35 5 5 .19 3 Elementary School Language Arts 1.50 12.00 90 4 6 -.50 3 Elementary School Mathematics 12.00 6.00 35 3 7 .09 4 Elementary School Mathematics 25.00 7.00 80 4 8a .31 7 Middle School Mathematics 18.00 6.00 40 4 8b .77 8 Middle School Mathematics 18.00 6.00 40 4 9 .96 8 Middle School Language Arts 13.00 24.00 95 5 10 -.29 9 High School Science 25.00 6.00 40 3 11 .45 1 Elementary School Language Arts 10.00 36.00 50 5 12 -1.65 1 Elementary School Mathematics 6.00 16.00 45 4 13 2.54 8 Middle School Science 20.00 9.00 50 3 14 .45 8 Middle School Mathematics 11.00 6.00 40 3 15 1.81 7 Middle School Social Studies 1.00 12.00 50 4 16 .39 8 Middle School Language Arts 6.00 12.00 65 3 17 .26 7 Middle School Mathematics 10.00 24.00 80 3 Final Report 83 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 18 -1.70 9-12 High School Foreign Languagea 18.00 30.00 98 4 19 .61 5 Elementary School Science 2.00 12.00 70 4 20 -.15 5 Elementary School Social Studies 3.00 12.00 20 3 21 1.11 2 Elementary School Mathematics 2.00 24.00 90 4 22 .96 4 Elementary School Science 10.00 .25 -- 2 23 1.04 1 Elementary School Language Arts 14.00 18.00 75 5 24 .04 2 Elementary School Mathematics 4.00 12.00 85 4 25 .14 1 Elementary School Language Arts 10.00 .25 50 4 26 .46 K Elementary School Mathematics 28.00 48.00 90 5 27 .82 5 Elementary School Language Arts 11.00 2.00 85 4 28 1.21 5 Elementary School Mathematics 8.00 -- 90 4 29 .22 4 Elementary School Social Studies 5.00 24.00 50 5 30 .59 K Elementary School Science 4.00 36.00 75 3 31 -.26 9-12 High School Foreign Languageb 9.00 36.00 90 5 32 .76 4 Elementary School Science 5.50 18.00 90 5 33 1.70 1 Elementary School Science 32.00 12.00 80 3 34 .53 7 Middle School Mathematics 7.50 12.00 90 4 35 .36 6 Middle School Mathematics 14.00 18.00 70 4 36 1.50 5 Elementary School Science 2.00 7.00 95 5 37 -.17 4 Elementary School Mathematics 4.00 4.00 95 3 Final Report 84 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 38 .17 8 Middle School Mathematics 13.00 4.00 90 3 39 -.72 2 Elementary School Science 18.00 4.00 55 4 40 .68 10-12 High School Mathematics 4.00 36.00 95 5 41 .14 3 Elementary School Mathematics 16.00 36.00 85 5 42 .55 2 Elementary School Language Arts 4.58 19.00 -- 4 43 .13 -- High School Language Arts 2.67 13.00 75 4 44 .10 -- Elementary School Language Arts 5.67 36.00 55 5 45 -.06 3 Elementary School Mathematics 20.58 43.00 60 4 a. Spanish. b. German. Figure C.2. Year One Evaluation Study Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 1 .21 10-12 High School Social Studies 10 5 20 3 2 1.22 6 Middle School Mathematics 7 24 50 4 3 .62 6 Middle School Social Studies 15 24 30 3 4 .06 8 Middle School Language Arts 12 6 50 3 5 -.13 7 Middle School Social Studies 30 24 35 2 6 .95 3 Elementary School Language Arts -- -- -- -- 7 .12 5 Elementary School Science -- -- -- -- 8 .48 5 Elementary School Mathematics 27 12 85 5 9 .65 5 Elementary School Social Studies 3 30 60 4 Final Report 85 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 10 .82 2 Elementary School Mathematics 20 36 60 4 11 1.18 2 Elementary School Mathematics 18 36 60 4 12 .57 1 Elementary School Language Arts 22 10 50 3 13a .38 2 Elementary School Language Arts 19 4 10 2 13b .17 1 Elementary School Language Arts -- -- -- -- 14 -.18 8 Middle School Mathematics 13 18 90 4 15 .57 8 Middle School Language Arts -- -- -- -- 16 .69 3 Elementary School Mathematics 23 48 65 4 17 .15 2 Elementary School Mathematics 17 36 80 5 18 .45 7 Middle School Mathematics 7 24 80 5 19 .48 6 Middle School Mathematics 10 12 75 4 20 -.83 6 Middle School Mathematics 11 6 60 4 21 .90 6 Middle School Mathematics 15 3 80 3 22 .16 6 Middle School Mathematics 8 5 75 4 23 .66 6 Middle School Mathematics 8 2 75 3 24a -.20 6 Middle School Mathematics 15 1 90 3 24b .08 6 Middle School Mathematics 15 1 90 3 25 1.32 6 Middle School Mathematics 11 18 75 4 26 -.13 6 Middle School Mathematics 13 7 60 4 27 .45 1 Elementary School Language Arts 7 12 -- 3 Final Report 86 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 28 .28 8 Middle School Mathematics -- -- -- -- 29 -.34 6 Middle School Computer Education -- -- -- -- 30 1.34 -- Elementary School Language Arts -- -- -- -- 31 .22 5 Elementary School Social Studies 4 18 75 5 32 .72 5 Elementary School Science 4 12 75 4 33 -.32 6 Middle School Language Arts 15 6 30 2.5 34 .39 5 Elementary School Science 16 24 70 4 35 .49 5 Elementary School Science -- -- -- -- 36 .08 5 Elementary School Social Studies 12 24 90 4 37 .20 3 Elementary School Language Arts 6 5 40 4 38 .27 1 Elementary School Language Arts 5 4 35 3 39 .03 6 Middle School Science 9 24 80 5 40 .66 7 Middle School Foreign Languageb 5 12 60 5 41 .83 -- Middle School -- -- -- -- -- 42 .13 12 High School Mathematics 8 36 85 4 43a .19 9 High School Science 9 36 85 5 43b .36 9 High School Science 9 36 85 5 43c .83 9 High School Science 9 36 85 5 44 1.23 9-12 High School Career Education 6 24 60 3 45 .60 9 High School Language Arts 5 24 90 4 Final Report 87 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 46 .57 11-12 High School Social Studies 12 36 80 4 47 .28 10-12 High School Mathematics 9 18 80 4 48 1.01 9-12 High School Music 15 36 40 4 49 -.06 7 Middle School Mathematics 7 18 85 5 50 1.66 6 Middle School Mathematics 4 12 95 5 51 1.03 3 Elementary School Social Studies 16 2 40 3.5 52 -.38 5 Elementary School Social Studies 32 3 40 3 53 .90 5 Elementary School Science 13 3 45 3 54 .78 5 Elementary School Science 25 3 80 3 55 .23 5 Elementary School Social Studies 15 2 40 2 56 .85 3-6 Elementary School Language Arts 11 3 75 4 57 -.55 9-12 High School Mathematics -- -- 50 -- 58 .45 9 High School Social Studies 3 24 70 4 59 .24 -- High School Mathematics 2 24 90 4 60 1.43 1 Elementary School Mathematics 5 24 75 5 61 -.83 7 Middle School Social Studies 23 4 40 4 62 -.05 2 Elementary School Social Studies 13 27 90 4 63 1.20 5 Elementary School Science 10 18 85 4 64 .61 10-12 High School Science 4 12 80 5 65 -.62 11-12 High School Mathematics 3 24 90 5 Final Report 88 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 Teacher Effect Size Grade Level School Level Content Area # Yrs Teach #Mos Tech % Class Time Confidence 66 1.54 10-11 High School Mathematics 23 24 80 5 67 -.50 8 Middle School Science 3 24 60 4 68 -.26 7 Middle School Science 5 36 85 5 69a .09 7 Middle School Language Arts 6 54 25 4 69b .25 7 Middle School Language Arts 6 54 25 4 69c -.50 7 Middle School Language Arts 6 54 25 4 70 1.33 6 Middle School Language Arts 13 60 80 5 71 .50 7 Middle School Language Arts 5 36 60 5 72 .96 8 Elementary School Social Studies -- -- -- -- 73 .15 7-8 Elementary School Science -- -- -- -- 74 .32 9 High School Social Studies -- -- -- -- 75 -.39 3 Elementary School Mathematics 7 3 25 3 76 -.02 4-5 Elementary School Mathematics 5 2 25 3 77 .14 K Elementary School Mathematics 14 3 20 3 78 -.24 4 Elementary School Language Arts 25 60 50 4 79 .11 4 Elementary School Language Arts -- -- -- -- a. French. Final Report 89 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010 References Borenstein, M., Hedges, L. V., Higgins, J., & Rothstein, H. (2005). Comprehensive meta-analysis (Version 2.2.050) [computer software]. Englewood, NJ: Biostat. Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to metaanalysis. West Sussex, England: John Wiley & Sons. Cohen, J. (1977). Statistical power for the behavioral sciences (Rev. ed.). New York: Academic Press. Cooper, H. (2009). Research synthesis and meta-analysis: A step-by-step approach (4th ed.). Thousand Oaks, CA: SAGE Publications. Glass, G. V. (1976). Primary, secondary, and meta-analyses of research. Educational Researcher, 5, 3-8. Harlow, L. L., Muliak, A. A., & Steiger, J. H. (Eds.). (1997). What if there were no significance tests? Mahwah, NJ: Erlbaum. Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic Press, Inc. Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Correcting error and bias in research findings (2nd ed.). Thousand Oaks, CA: SAGE Publications. Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: SAGE Publications. Lou, Y., Abrami, P. C., Spence, J. C., Poulsen, C., Chambers, B., & d’Apollonia, S. (1996). Within-class grouping: A meta-analysis. Review of Educational Research, 66 (4), 423-458. Marzano, R. J., & Haystead, M. W. (2009). Final report: An evaluation study of the effects of Promethean ActivClassroom on student achievement. Englewood, CO: Marzano Research Laboratory. Murphy, K. R., & Myors, B. (2004). Statistical power analysis: A simple and general model for traditional and modern hypothesis tests (2nd ed.). Mahwah, NJ: Erlbaum. Osborne, J. W. (2003). Effect sizes and disattenuation of correlation and regression coefficients: Lessons from educational psychology. Practical Assessment, Research and Evaluation, 8 (11). Retrieved March 20, 2009, from http://PAREonline.net/getvn.asp?v=8&n=11. SPSS. (2009). PASW® Statistics (Version 17.0.2) [computer software]. Chicago, IL: SPSS. Wilson, D. B. (2005). Meta-analysis macros for SAS, SPSS, and Stata (Version 2005.05.23) [computer software]. Retrieved June 12, 2009, from http://mason.gmu.edu/ ~dwilsonb/ma.html. Final Report 90 2009/2010 Second Year Promethean Evaluation Study Marzano Research Laboratory December 2010