COMPASS ANALYSIS Executive Summary A random sample of students was gathered (n=244) to understand what explains student success in students' initial math courses at ISU. Analysis of the data suggests that passing remedial and non-remedial courses is most likely influenced by students' scores on the second COMPASS sub scores and on the number of college preparatory mathematics courses successfully completed in high school. When non-remedial courses are analyzed (n=54), SAT math scores and the number of college preparatory courses successfully completed in high school explain grades in initial ISU math courses; the higher the SAT math score and the higher number of college preparatory courses successfully completed, the higher the grades in students' initial math courses. Further analysis indicates that the latter of these, number of college preparatory courses taken in high school, is the most significant factor. Factors Explaining Success in Initial Math Courses at ISU Background The University Academic Advisement Committee (UAAC) requested that a study be conducted so its members could better understand what different factors--SAT math scores, COMPASS test scores, and high school preparation for collegiate math courses in particular--explain success in collegiate math courses. UAAC requested that Assessment and Accreditation, as part of the Division of Academic Affairs, conduct the study. Methodology Sample A random sample of 305 students taking courses in 200505 was taken, of which 244 contained sufficient information for analysis. Data A request was submitted to the registrar's office and the Student Affairs Office of the College of Arts and Sciences to gather numerous amounts of information. The following information was received: • • • • • • The students' university 991 number to control for multiple submissions of information The initial math course taken by each student The grade in the students' initial math course, which was then transformed into the appropriate grade point defined in the university catalog. For most purposes, this served as the dependent variable for statistical purposes Year of high school graduation; this was used as a possible control for the argument that number of years elapsed since high school graduate could be a factor in initial math grades Term admitted to ISU ACT and SAT math and English scores • The students' Compass scores (all sub-scores, if applicable, were supplied) In addition, the high school transcripts for each student were supplied, allowing us to gather the following information: • Of high school math courses (pre-algebra, algebra, geometry, algebra 2, precalculus/trigonometry, calculus/AP calculus), how many such courses were taken Dependent Variables The subject of study is student success in initial math courses taken at ISU. Because students take either remedial or non-remedial courses, and because remedial courses in particular are grades on a bimodal scale, "S" meaning satisfactory and "U" meaning unsatisfactory, it was necessary to conduct three separate studies: • • • A study of success in remedial courses, using "S" denoting success, and "U" denoting lack of success A study of student grades in non-remedial courses, using grade points as the measure of those grades A study of success (passing grades) and non-success (non-passing grades--Ds, Fs, or W/WF's) in all courses Independent Variables The following were used as independent, or explanatory, variables for this analysis: • Year of high school graduation (to control for time) • SAT math scores • COMPASS math placement scores1, in particular the first pre-algebra test and the second algebra test • High school college preparatory math courses taken (identified as "1" for course taken, and "0" for course not taken); by "college-preparatory" we mean prealgebra, algebra 1, algebra 2, geometry, pre-calculus/trigonometry, calculus/AP calculus. • Number of college preparatory math courses not failed in the list above • Number of college preparatory math courses in which students earned grades of C or higher • Number of years between high school graduation and initial math course taken 1 The COMPASS placement test is designed by ACT, Inc, “to help accurately place students in college math courses…it indicates how well [a student] know(s) the background material needed to be successful in…college level math course(s).”1 It is used by the mathematics department and College of Arts and Sciences at Indiana State University as an advising tool to assist students in maximizing the chances for success in initial registration in math courses. The COMPASS test is divided into four specific tests, the first two of which are of special importance in respect to assigning students to either remedial or non-remedial mathematics courses. The first test is a prealgebra test. Students who score below 44 on this test are assigned to Math 010, and those who score 44 or higher on this test are assigned to Math 011. If students are successful enough on the pre-algebra test, ACT enables students to proceed toward completion of the second test, which tests them on algebra. Students who score between 1 and 40 on this test are encouraged to take Math 011 (with advisor discretion), and those who score higher than 40 are encouraged to take non-remedial math. Analytical Strategy We use frequency statistics and cross-tabulations to present the data in simple form. Second, in order to observe what factors explain success in remedial math courses, a logistic regression was performed that tests for the probability of success, using (1) "S" grades in remedial courses, or (0) "U" grades in remedial courses as the dependent variable. We used the above independent variables to accomplish our purpose. Third, in order to observe what factors explain success in remedial and non-remedial courses, a logistic regression was also performed, using (1) as passing grades ("S" in remedial courses and A, B, and C grades in non-remedial courses) and (0) for nonpassing grades ("U" in remedial courses and D, F, W, and WF). Fourth, for non-remedial courses, we used the grade number assigned to each course (i.e, 4.0=A, 3.5=B+, etc.) as the dependent variable, and the above independent variables, in tests of association and in a regression analysis. In what follows, we attempt to answer each of the above research questions. Analysis Frequency Statistics The following table reveals basic frequency statistics for the variables included in this analysis: Frequency Statistics N Valid Mean Missing Median Std. Deviation Minimum Maximum Initial Math Course Grade Number 231 24 2.1450 2.0000 1.45122 .00 4.00 Initial Remedial Math 244 11 .05 .00 .225 0 1 Years Since HS Graduation 238 17 1.9328 .0000 5.01682 .00 30.00 238 17 2.7815 1.0000 5.02090 -2.00 31.00 238 17 2001.79 2003.00 5.064 1972 2005 62.95 65.00 18.380 17 98 Years Between Initial Math Course and HS Graduation Year Graduated From High School Compass Test 1 97 158 Compass Test 2 70 185 52.09 50.50 21.447 15 93 SAT Math Score 181 74 482.82 470.00 83.880 260 720 Total Number of Math Courses in HS 164 91 4.45 5.00 .980 0 6 164 91 4.24 4.00 1.187 0 6 Total Number of Math Courses in HS D Grades Not Included Initial Math Course Pass 244 11 .65 1.00 .477 0 Table One indicates that 35% of students in the sample fail their initial math course. Very few students in this sample spent much time out of college between high school graduation and initial matriculation into college, and students tended to wait one to three years between their last math courses taken in high school and their initial math course. Scores on the first COMPASS sub-test were fairly high, although this is no surprise since only very few students took remedial math. Scores on the second COMPASS math test were a bit lower. Students took between four and five college preparatory courses in high school, meaning that the average student completed algebra 2 in high school.2 This number appears to decline somewhat when D grades exclude students from this designation. 65% of students taking their initial math course pass. Table Two below reveals how many students passed and failed each of their respective initial math courses: Passing Initial Math Courses Initial Math Course Pass Did Not Pass Initial Math Course Passed Initial Math Course Initial Math Course Total Total MA010 0 2 2 MA011 1 10 11 MA102 44 81 125 MA 111 23 36 59 MA 112 2 3 5 MA 115 8 16 24 MA 131 7 11 18 85 159 244 For this specific sample of students, it appears that, with the exception of math remedial courses, thirty percent or more of students fail their initial math courses. Remedial Math Courses Table Three found at the end of this report compares demographic and test score information of remedial students with non-remedial students: In respect to background in math, there appear to be no surprises. For students taking remedial courses, SAT and COMPASS tests tend to be lower, as well as the number of college-preparatory math courses taken in high school--students taking remedial courses in their initial math experience tended on average to have completed geometry compared 2 1=pre-algebra, 2=algebra 1, 3=geometry, 4=algebra 2, 5=pre-calculus/trig, 6=calculus/AP calculus 1 to algebra 2 for those taking non-remedial courses as their initial collegiate math course at ISU. Table Four uses nonparametric methods to observe what best correlates with placement in remedial math courses: The higher the SAT math score, the higher the Compass score on test one (MCP1), the more years between initial math courses and high school graduation, and the total number of college preparatory courses taken in high school, the lower the probability that students will initially enroll in remedial courses work (which is clearly not a surprise). Explaining Factors Placing Students in Remedial Math Courses Years Between Initial Math Course and HS Graduation Mann-Whitney U Wilcoxon W Total Number of Total Number Math Courses in Year Graduated HS D Grades Not of Math From High School Compass Test 1 SAT Math Score Courses in HS Included 994.000 1462.000 247.500 455.000 526.500 641.000 1085.000 26887.000 325.500 521.000 617.500 732.000 -1.997 -.002 -2.877 -2.852 -2.967 -2.179 .046 .998 .004 .004 .003 .029 Z Asymp. Sig. (2tailed) a Grouping Variable: Initial Remedial Math NOTE: Because only one student in the sample failed his/her initial remedial course, a study of what explains success cannot be conducted. A prior non-randomized study performed in 2004, however, indicated that the COMPASS test, particularly the first subtest, was a good predictor of success in remedial courses. Success in Initial Math Courses Taken at ISU A comparison of those who passed or did not pass their initial math courses can be found in Table Five at the end of this report. The frequency data in the table indicate that there are differences, especially in respect to the COMPASS tests (first and second subtests) and the SAT math score. An initial logistic regression study indicates that there is a high likelihood that the second sub-test of the COMPASS, and number of courses taken in high school explain the probability of success in initial math courses. A stricter logistic regression analysis, much like stepwise regression, indicates that the most significant of these variables is number of courses taken in high school, especially when courses are counted that exclude D grades earned in such courses. Explaining Grades in Non-Remedial Courses Observing possible correlations between the students' success in their initial math course at ISU and the SAT math subtest, COMPASS subtests, and high school preparation for collegiate math courses, is informative because it is easier to determine which are significantly associated, thereby allowing us to better understand which can be good predictors. It is clear from the above table that SAT math scores, total number of math courses, and the COMPASS 2 test are significantly associated with higher grades in students' initial math courses at ISU.3 The higher the scores on these indicators, the higher the grade students' initial math courses. A stepwise regression analysis enables us to enter these variables in one equation, and focus only on those variables that are both significantly associated with grades in initial math courses and that independently explain that association (stepwise regression). A regressions analysis that includes these variables in such an equation, found at the end of this report, indicates that number of college preparatory math courses taken in high school and SAT math scores explain success in non-remedial initial math courses taken by students at ISU. Summary From the above analysis, the following appear to be reasonable statements: 1. When data for students taking remedial courses versus those taking non-remedial courses are compared, the latter tend to have higher SAT scores, higher COMPASS sub scores, and a higher number of college preparatory courses in high school. 2. The random sample used in this study indicates that 30% or more of all students taking non-remedial courses fail their initial math course at ISU 3. Higher scores on the second COMPASS subtest and number of college preparatory courses taken in high school suggest a higher probability that students will succeed in their initial math course taken at ISU (the sample includes students taking both remedial and non-remedial courses). A more detailed logistic regression, however, indicates that the most significant of these variables is number of courses taken in high school, especially when courses are counted that exclude D grades earned in such courses 4. A regressions analysis that includes SAT math scores, total number of math courses, and COMPASS 2 test scores , found at the end of this report, indicates that number of college preparatory math courses taken in high school and SAT math scores explain success in non-remedial initial math courses taken by students at ISU. One concern for these findings is that the model generated by the stepwise regression analysis indicates that there is a lower probability of students' passing their initial math courses beyond what can be explained through this analysis. For example, according to 3 Compass test 1 was not significant at 0.082 the regression analysis, a student with a 483 on the SAT math test (the average according to the frequency statistics) who took precalculus/trigonometry in high school would receive 2.3 grade points (a C) in math. This may be comforting in respect to supporting the notion that the analysis is close to the average grade points of all students taking nonremedial math courses initially (see frequency table at beginning of the report), but not so comforting in that such a grade appears low for a student with a strong pre-collegiate background. A number of reasons for this might be subjects of discussion for UAAC, such as: 1. Given that 30% of students in this sample failed their initial math course (excluding students who took remedial courses as their initial math course), is there a reason why the failure rate is this high? Or, is this failure rate what is to be expected? 2. Is the level of high school preparation in college preparatory math courses adequate? Group Statistics Initial Remedial Math Mean Std. Deviation Valid N (listwise) Unweighted Initial Math Course Was NonRemedial Initial Math Course Was Remedial Total Years Since HS Graduation .0147 .12127 68 68.000 Years Between Initial Math Course and HS Graduation .6471 .80604 68 68.000 Compass Test 1 66.0000 16.91683 68 68.000 SAT Math Score 503.6765 72.70682 68 68.000 Total Number of Math Courses in HS 4.7206 .87836 68 68.000 Total Number of Math Courses in HS D Grades Not Included 4.5294 1.07162 68 68.000 Years Since HS Graduation .1000 .31623 10 10.000 Years Between Initial Math Course and HS Graduation .9000 .73786 10 10.000 Compass Test 1 49.6000 15.07168 10 10.000 SAT Math Score 422.0000 59.02918 10 10.000 Total Number of Math Courses in HS 3.7000 .67495 10 10.000 Total Number of Math Courses in HS D Grades Not Included 3.6000 .84327 10 10.000 Years Since HS Graduation .0256 .15908 78 78.000 Years Between Initial Math Course and HS Graduation .6795 .79762 78 78.000 Compass Test 1 63.8974 17.49329 78 78.000 SAT Math Score 493.2051 75.90995 78 78.000 Total Number of Math Courses in HS 4.5897 .91787 78 78.000 Total Number of Math Courses in HS D Grades Not Included 4.4103 1.08635 78 78.000 Weighted Group Statistics Initial Math Course Pass Mean Std. Deviation Valid N (listwise) Unweighted Did Not Pass Initial Math Course Years Since HS Graduation .0000 .00000 13 13.000 Years Between Initial Math Course and HS Graduation .9231 .49355 13 13.000 Compass Test 1 69.9231 9.05043 13 13.000 SAT Math Score 489.2308 37.74068 13 13.000 Total Number of Math Courses in HS 4.2308 1.09193 13 13.000 Total Number of Math Courses in HS D Grades Not Included 3.7692 1.36344 13 13.000 Compass Test 2 Passed Initial Math Course 41.0000 19.23972 13 13.000 Years Since HS Graduation .0250 .15811 40 40.000 Years Between Initial Math Course and HS Graduation .6000 .74421 40 40.000 Compass Test 1 73.8500 11.30135 40 40.000 SAT Math Score 517.2500 68.19984 40 40.000 Total Number of Math Courses in HS 4.8750 .79057 40 40.000 Total Number of Math Courses in HS D Grades Not Included 4.8000 .79097 40 40.000 Compass Test 2 Total 52.8750 20.78977 40 40.000 Years Since HS Graduation .0189 .13736 53 53.000 Years Between Initial Math Course and HS Graduation .6792 .70092 53 53.000 Compass Test 1 72.8868 10.84446 53 53.000 SAT Math Score 510.3774 62.97015 53 53.000 Total Number of Math Courses in HS 4.7170 .90676 53 53.000 Total Number of Math Courses in HS D Grades Not Included 4.5472 1.04819 53 53.000 49.9623 20.88518 53 53.000 Compass Test 2 Weighted Correlations Initial Math Course Grade Number Pearson Correlation Initial Math Course Grade Number SAT Math Score 1.000 .371 .616 .389 .340 .819 .371 1.000 .215 .510 .117 .231 .616 .215 1.000 .411 .232 .504 Compass Test 2 .389 .510 .411 1.000 .202 .328 .340 .117 .232 .202 1.000 .350 .819 .231 .504 .328 .350 1.000 Initial Math Course Grade Number SAT Math Score . .003 .000 .002 .006 .000 .003 . .059 .000 .199 .047 Total Number of Math Courses in HS D Grades Not Included .000 .059 . .001 .046 .000 Compass Test 2 .002 .000 .001 . .072 .008 .006 .199 .046 .072 . .005 .000 .047 .000 .008 .005 . 54 54 54 54 54 54 Year Graduated From High School Initial Math Course Pass N Initial Math Course Pass Total Number of Math Courses in HS D Grades Not Included Year Graduated From High School Initial Math Course Pass Sig. (1-tailed) Total Number of Math Courses in HS D Grades Not Year Graduated SAT Math Score Included Compass Test 2 From High School Initial Math Course Grade Number SAT Math Score 54 54 54 54 54 54 Total Number of Math Courses in HS D Grades Not Included 54 54 54 54 54 54 Compass Test 2 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 Year Graduated From High School Initial Math Course Pass Model Summary(c) Change Statistics Model R R Square Adjusted R Square Std. Error of the Estimate R Square Change F Change df1 df2 Sig. F Change 1 .616(a) .380 .368 1.11531 .380 31.874 1 52 .000 2 .663(b) .440 .418 1.07058 .060 5.437 1 51 .024 a Predictors: (Constant), Total Number of Math Courses in HS D Grades Not Included b Predictors: (Constant), Total Number of Math Courses in HS D Grades Not Included, SAT Math Score c Dependent Variable: Initial Math Course Grade Number ANOVA(c) Model 1 Sum of Squares Regression Residual Total 2 df Mean Square 39.649 1 39.649 64.684 52 1.244 104.333 53 Regression 45.881 2 22.940 Residual 58.453 51 1.146 104.333 53 Total F Sig. 31.874 .000(a) 20.015 .000(b) a Predictors: (Constant), Total Number of Math Courses in HS D Grades Not Included b Predictors: (Constant), Total Number of Math Courses in HS D Grades Not Included, SAT Math Score c Dependent Variable: Initial Math Course Grade Number Durbin-Watson 1.695 Coefficients(a) Model 1 Unstandardized Coefficients (Constant) Total Number of Math Courses in HS D Grades Not Included SAT Math Score 2 (Constant) Total Number of Math Courses in HS D Grades Not Included SAT Math Score Standardized Coefficients -1.143 .644 .769 .136 -3.643 1.237 .702 .134 .005 .002 -1.776 .082 5.646 .000 -2.944 .005 .563 5.241 .000 .250 2.332 .024 .616 a Dependent Variable: Initial Math Course Grade Number Excluded Variables(c) Model Tolerance 1 2 SAT Math Score .250(a) 2.332 .024 .310 .954 Compass Test 2 .163(a) 1.376 .175 .189 .831 .208(a) 1.901 .063 .257 .946 Compass Test 2 .047(b) .357 .722 .050 .644 Year Graduated From High School .191(b) 1.812 .076 .248 .941 Year Graduated From High School SAT Math Score a Predictors in the Model: (Constant), Total Number of Math Courses in HS D Grades Not Included b Predictors in the Model: (Constant), Total Number of Math Courses in HS D Grades Not Included, SAT Math Score c Dependent Variable: Initial Math Course Grade Number