Speaker: Ryan Barnhart, York University
Department of Psychology
Title: Repeated Measurement Designs for Pre-Post Analysis, Difference Scores and ANOVA for fixed and unfixed measurement occasions: It is not the mean difference but the mean slope, time matters in these models
Abstract: On many occasions the adage “The Simplest Solution is the Best Solution” is a very true statement. However, the pitfalls of assumed simple designs can actually be many and lead many a practitioner awry if not properly considering all the observed and unobserved elements within formal equations. This talk focuses on one such flaw that I have observed repeatedly in use in repeated measurements designs, where the unobserved or unaccounted for effect of time is often neglected or overlooked when estimating mean differences/treatment effects.
I will first demonstrate the latent assumptions regarding time in the equations of the paired t-test, the testing of difference scores and 2-occasion repeated measures ANOVA for fixed measurement occasions. Second, I will provide proof for and demonstrate the presence of biases that occur when random and non-random differences in the measurement occasions are present.
Suggested Readings: TBA