Title of Resource Activity: Conducting a Repeated Measures ANOVA Author(s) Natalie Ciarocco & Jaclyn Racaniello Institution Monmouth University This activity has students conduct a repeated measures ANOVA using data from a research scenario. Students determine the statistical Brief Description: analysis to use, identify the independent and dependent variables, set up a data file, enter the provided data, run the appropriate analysis, and summarize the results. Repeated Measures ANOVA; PASW (SPSS) Calculation; Identifying IVs Keywords: & DVs; Interpretation of Output Author Contact nciarocc@monmouth.edu Information: Additional Information: TeachPsychScience.org is made possible with grant support from the American Psychological Society (APS) Fund for Teaching and Public Understanding of Psychological Science to the site creators Gary Lewandowski, Natalie Ciarocco, and David Strohmetz. All materials on this site have been subjected to a peer review process. We welcome additional resources (www.teachpsychscience.org/submissions). © 2010 Natalie Ciarocco and Jaclyn Racaniello. All Rights Reserved. This material may be used for noncommercial educational purposes. All other uses require the written consent of the authors. Instructors: In this activity, students will take data from a fictitious design to practice conducting a repeated measures ANOVA. First, provide students with the research scenario and the accompanying questions to have them determine the research design, statistical analysis to use, and independent and dependent variables. Next, have students set up a data file using the included data from the study and run the appropriate analysis. Students should summarize their results. An answer key and output are included. Although PASW (SPSS) output is provided, this activity could be run using any statistics program or students could perform calculations by hand. Research Scenario A researcher is interested in testing which brand of coffee tastes better. Ten people were selected to participate in tasting four different types of coffee: Maxwell House, Folgers, Starbucks, and Dunkin Donuts. All coffee brands were made by the researcher from fresh grinds and the order in which each participant tasted the coffees was counterbalanced. After tasting each brand of coffee, participants rated the quality of taste on a 7-point scale (1 = terrible to 7 = delicious). It is hypothesized that the brands will differ in quality of taste. 1. What type of methodological design is being utilized? 2. What statistical analysis should be utilized to test the prediction? 3. Identify the independent variable(s): 4. Identify the dependent variable(s): 5. Set up an appropriate data file and enter the provided data. 6. Run the appropriate analysis and summarize your findings below: _____ ______ Participant #: 1 Age: ,;{S ~e mCl Gender: Ie.. Dunkin Donuts score out of7: L\ Folgers score out of7: 1 \ Starbucks score out of 7: Maxwell score out of7: S - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- Participant #: 2 Age: ~ 1 Gender: rn C/k: Dunkin Donuts score out of 7: L1 Folgers score out of7: lo Starbucks score out of7:.J Maxwell score out of 7: \.p - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - -- - --- - -- - - - - - - -- Participant #: 3 Age: 36 Gender: ~malt' Dunkin Donuts score out of 7:Lj Folgers score out of7: (p Starbucks score out of 7: I Maxwell score out of7: (Q - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- Participant #: 4 Age: ;<0, Gender: +cmo k: Dunkin Donuts score out of 7: S Folgers score out of7: lP Starbucks score out of7: 3 Maxwell score out of 7: 5 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - -- - - - - - - - - - - - - -Participant #: 5 Age: \~ Gender: ~mo. \.e Dunkin Donuts score out of 7: (p Folgers score out of7: ':) Starbucks score out of7: L-I Maxwell score out of7: 1-) - - - - - -- - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - -- Participant #: 6 Age: 31 Gender: pi u\e Dunkin Donuts score out of7: S Folgers score out of7: ~ Starbucks score out of7: \ Maxwell score out of 7: 4 - - - - - - - - - - - - - - - -- - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- Participant #: 7 Age: L\0 Gender: ~e /YO Ie / Dunkin Donuts score out of 7: ;) Folgers score out of7: -, Starbucks score out of7: JMaxwell score out of 7: Lv - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - -Participant #: 8 Age: S\ Gender: )'YIQ \e. Dunkin Donuts score out of7: 3 Folgers score out of7: I Starbucks score out of7: :;l. Maxwell score out of 7: c:) - - - - - - - - - - - - - - - -- - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - -Participant #: 9 Age: L\d c YnCt l-e Gender: +' Dunkin Donuts score out of 7: Folgers score out of 7: ~ {p Starbucks score out of7: J Maxwell score out of 7: S - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - -- - -- Participant #: 10 Age: (06 Gender: n1C\ 1<: Dunkin Donuts score out of7: F olgers score out of 7: 1.0 Starbucks score out of7: L\ Maxwell score out of7: Y 5 ANSWER KEY Conducting a One-way ANOVA Research Scenario A researcher is interested in testing which brand of coffee tastes better. Ten people were selected to participate in tasting four different types of coffee: Maxwell House, Folgers, Starbucks, and Dunkin Donuts. All coffee brands were made by the researcher from fresh grinds and the order in which each participant tasted the coffees was counterbalanced. After tasting each brand of coffee, participants rated the quality of taste on a 7-point scale (1 = terrible to 7 = delicious). It is hypothesized that the brands will differ in quality of taste. 1. What type of methodological design is being utilized? multi-group design (repeated measures) 2. What statistical analysis should be utilized to test the prediction? 3. Identify the independent variable(s): Repeated Measures ANOVA brand of coffee 4. Identify the levels of each independent variable: Maxwell House, Folgers, Starbucks, and Dunkin Donuts 5. Identify the dependent variables: quality of taste (range 1-7) 6. Set up an appropriate data file and enter the provided data. 7. Run the appropriate analysis and summarize your findings below: The hypothesis was supported. The quality of taste was rated as significantly different between brands. Pairwise comparisons revealed that the ratings of Folgers (M = 6.20) coffee were significantly higher than the ratings of Maxwell House (M = 5.00) , Starbucks (M = 2.30), and Dunkin Donuts (M = 4.20). Similarly, Starbucks was rated significantly lower than each of the other brands. PASW (SPSS) Output Conducting a Repeated Measures ANOVA General Linear Model Within-Subjects Factors Descriptive Statistics Measure:MEASURE_1 coffee_1 Mean Dependent Variable Std. Deviation 1 Dunkin_Donuts Dunkin_Donuts 4.2000 2 Folgers Folgers 3 Starbucks Starbucks Maxwell Maxwell 4 N 1.13529 10 6.2000 .63246 10 2.3000 1.15950 10 5.0000 .81650 10 Multivariate Testsb Effect coffee_1 Value F Hypothesis df a Error df Sig. Pillai's Trace .904 21.953 3.000 7.000 .001 Wilks' Lambda .096 21.953a 3.000 7.000 .001 9.409 a 3.000 7.000 .001 a 3.000 7.000 .001 Hotelling's Trace Roy's Largest Root 21.953 9.409 21.953 a. Exact statistic b. Design: Intercept Within Subjects Design: coffee_1 Mauchly's Test of Sphericityb Measure:MEASURE_1 Within Subjects Effect coffee_1 Epsilona Mauchly's W .348 Approx. Chi-Square 8.151 df GreenhouseGeisser Sig. 5 .151 .609 Huynh-Feldt Lower-bound .756 Tests the null hypothesis that the error covariance matrix of the orthonormalized transformed dependent variables is proportional to an identity matrix. a. May be used to adjust the degrees of freedom for the averaged tests of significance. Corrected tests are displayed in the Tests of WithinSubjects Effects table. b. Design: Intercept Within Subjects Design: coffee_1 .333 Tests of Within-Subjects Effects Measure:MEASURE_1 Type III Sum of Squares Source coffee_1 Error(coffee_1) df Mean Square F Sphericity Assumed 80.475 3 26.825 23.535 .000 Greenhouse-Geisser 80.475 1.827 44.048 23.535 .000 Huynh-Feldt 80.475 2.268 35.480 23.535 .000 Lower-bound 80.475 1.000 80.475 23.535 .001 Sphericity Assumed 30.775 27 1.140 Greenhouse-Geisser 30.775 16.443 1.872 Huynh-Feldt 30.775 20.414 1.508 Lower-bound 30.775 9.000 3.419 Tests of Within-Subjects Contrasts Measure:MEASURE_1 Type III Sum of Squares Source coffee_1 coffee_1 Linear 1.125 1 1.125 1.074 .327 Quadratic 1.225 1 1.225 2.436 .153 Cubic 78.125 1 78.125 41.790 .000 Linear 9.425 9 1.047 Quadratic 4.525 9 .503 16.825 9 1.869 Error(coffee_1) Cubic df Mean Square F Tests of Between-Subjects Effects Measure:MEASURE_1 Transformed Variable:Average Type III Sum of Squares Source Intercept df Mean Square 783.225 1 783.225 2.525 9 .281 Error Estimated Marginal Means 1. Grand Mean Measure:MEASURE_1 95% Confidence Interval Mean 4.425 Sig. Std. Error .084 Lower Bound 4.236 Upper Bound 4.614 F 2791.693 Sig. .000 Sig. 2. coffee_1 Estimates Measure:MEASURE_1 95% Confidence Interval coffee_1 Mean Std. Error Lower Bound Upper Bound 1 4.200 .359 3.388 5.012 2 6.200 .200 5.748 6.652 3 2.300 .367 1.471 3.129 4 5.000 .258 4.416 5.584 Pairwise Comparisons Measure:MEASURE_1 (I) (J) coffee_1 coffee_1 1 3 4 Sig.a Std. Error Lower Bound Upper Bound * .537 .005 -3.216 -.784 * 1.900 .379 .001 1.044 2.756 4 -.800 .573 .196 -2.097 .497 1 2.000* .537 .005 .784 3.216 3 3.900* .504 .000 2.759 5.041 4 * 1.200 .249 .001 .636 1.764 1 -1.900* .379 .001 -2.756 -1.044 2 -3.900* .504 .000 -5.041 -2.759 4 * -2.700 .539 .001 -3.918 -1.482 1 .800 .573 .196 -.497 2.097 2 * .249 .001 -1.764 -.636 * .539 .001 1.482 3.918 2 -2.000 3 2 95% Confidence Interval for Differencea Mean Difference (IJ) -1.200 3 2.700 Based on estimated marginal means *. The mean difference is significant at the .05 level. a. Adjustment for multiple comparisons: Least Significant Difference (equivalent to no adjustments). Multivariate Tests Value Pillai's trace Wilks' lambda .904 F Hypothesis df Error df Sig. a 3.000 7.000 .001 a 21.953 .096 21.953 3.000 7.000 .001 Hotelling's trace 9.409 21.953a 3.000 7.000 .001 Roy's largest root 9.409 21.953a 3.000 7.000 .001 Each F tests the multivariate effect of coffee_1. These tests are based on the linearly independent pairwise comparisons among the estimated marginal means. a. Exact statistic