Presentation on theme: "Using a Repeated Measures ANOVA to Analyze the Data from a Pretest- Posttest Design: A Potentially Confusing Task Schuyler Huck and Robert McLean."— Presentation transcript:
Using a Repeated Measures ANOVA to Analyze the Data from a Pretest- Posttest Design: A Potentially Confusing Task Schuyler Huck and Robert McLean
Overview Pretest-Posttest Design Repeated Measures ANOVA Gain Scores Problems with Repeated Measures ANOVA Conservative F value Interaction Effect versus Gain Scores Post-Hoc Tests Advantages of Gain Scores over Repeated Measures
SubjectPretestPost- test Experi- mental Group X 111 X 112 X 113 X 121 X 122 X 123 Control Group X 114 X 115 X 116 X 124 X 125 X 126 Repeated Measures ANOVA- Effect of treatment Effect of time Interaction between treatment & time Gain Scores Difference between pretest and posttest scores for each person Posttest-pretest Use a one-way ANOVA, main effect of treatment
Problems with Repeated Measures ANOVA SubjectPretestPost- test Experi- mental Group X 111 X 112 X 113 X 121 X 122 X 123 Control Group X 114 X 115 X 116 X 124 X 125 X 126 F value is too small- pretest scores are included in effect of treatment, but no subject experienced treatment before the pretest scores. Treatment effects only influence posttest data. Interaction effect is true main effect of treatment- the interaction examines the difference between groups depending upon pretest versus posttest scores. Post Hoc Problems- Simple main effect tests run the risk of a “type IV error” and alpha values are controversial. Multiple comparison t-tests make sense only if gain score analyses was used.
Gain Scores vs Repeated Measures ANOVA F ratios of Repeated Measures ANOVA are not useful F main effect treatment- not an accurate estimate of treatment effect. F interaction is equivalent to gain scores. F main effect of time- no true experimental value. Gain scores Equivalent information as ANOVA, but without the confusion and controversy. Principle of parsimony- simpler is better!