Detecting Instruction Effects—Deciding Between Covariance Analytical and Change-Score Approach

Abstract The article focuses on estimating effects in nonrandomized studies with two outcome measurement occasions and one predictor variable. Given such a design, the analysis approach can be to include at previous time point as regression model (ANCOVA), or predict change-score of variable (CHANGE). Researchers demonstrated that both approaches result different conclusions regarding reported effect. Current recommendations when apply which are, part, contradictory. In addition, they lack direct reference educational instructional research contexts, since do not consider latent models variables are measured without error. This contribution assists researchers making decisions their model. Using an underlying hypothetical data-generating model, we identify for kind scenario (i.e., under assumptions) defined true effect equals estimated coefficients ANCOVA CHANGE approach. We give empirical examples from discuss is more appropriate, respectively.