General Growth Modeling in Experimental Designs: A Latent Variable Framework for Analysis and Power Estimation

Researchers frequently need to consider individual student growth differences over time. In this report, researchers Bengt Muthen and Patrick Curran explore new types of statistical growth models and the power of detecting various treatment effects through a latent variable framework. Analyzing treatment effects using artificial data, the authors demonstrate the importance of going beyond the analysis of covariance approach to the use of more than two timepoints. They also show that interaction effects can be detected without unduly large sample sizes if the interaction effects are sizable. Analyzing data from a behavior modification program for young children, the researchers demonstrated how growth modeling can reveal different and more in-depth information about treatment effects.