A Comparison of Five Statistical Methods for Analyzing Pretest-Posttest Designs

Abstract

Five methods for analyzing data arising from research involving pretests and posttests are considered. These methods include: (1) posttest analysis only; (2) analysis of raw gain scores (posttest minus pretest); (3) analysis of the data with a pretest-posttest factor included in the statistical model; (4) analysis of posttest data with pretests as a covariate, and (5) analysis of gain scores with pretests as a covariate. The characteristics of each are discussed, with a conclusion that the fifth method is superior to the others when the assumptions underlying covariance analysis are met.