Evaluating Pretest-Posttest Control Group Experiments

Abstract

The analysis of pretest-posttest control group experiments was simulated using gain scores, analysis of covariance, and posttest only methods. In addition, a sequential procedure was evaluated in which gain scores were used unless the initial difference between the groups on the pretest was large at which time a covariance was used. The sequential strategy was conservative with regard to Type I error. The power of the methods with varying correlations between pre- and posttest and varying treatment effects favored a standard analysis of covariance.