ABSTRACT
Analysis of covariance (ANCOVA) is a statistical technique that combines regression analysis and analysis of variance. It can be helpful in nonrandomized studies in drawing more accurate conclusions. However, precautions have to be taken, otherwise analysis of covariance can be misleading in some cases. In this chapter we indicate what the purposes of ANCOVA are, when it is most effective, when the interpretation of results from ANCOVA is “cleanest,” and when ANCOVA should not be used. We start with the simplest case, one dependent variable and one covariate, with which many readers may be somewhat familiar. Then we consider one dependent variable and several covariates, where our previous study of multiple regression is helpful. Multivariate analysis of covariance (MANCOVA) is then considered, where there are several dependent variables and several covariates. We show how to run MANCOVA on SAS and SPSS, interpret analysis results, and provide a guide for analysis.
