Linear Versus Nonlinear Analysis in the Measurement of Effects in a Quasi-Experimental Design

Abstract

In this study, a two-step procedure for the analysis of pretest-posttest data is developed and illustrated. In the first step, a nonlinear canonical correlation analysis was conducted on data from a pretest-posttest control group design. This technique transforms the measure of the pretests (including the scores on the treatment variable) and the posttests in such a way that they become interrelated linearly and possible deviations from the linear model due to nonlinearity have been minimized. In the second step, the resulting optimally scaled set of pretest and posttest measures were analyzed using covariance procedures to assess program effects. The resultant variance appeared to be increased substantially, either by a better prediction of posttest scores from pretest scores, by a better estimation of the effect of the treatment, or by both. It is concluded that the two-step procedure indeed has important advantages.