Important considerations in using statistical procedures to control for nuisance variables in non-experimental studies

In this paper, it is argued that the effects of statistically controlling for confounding variables in non-experimental studies have received insufficient attention. One of the reasons for this inattention is likely because many researchers do not fully appreciate what a statistical control strategy entails and how the use of such a strategy can affect the hypotheses actually tested in a study. In order to address these issues, this article (a) considers what the use of statistical control entails computationally, (b) reviews studies in which researchers have focused on the unique variance added by a predictor in testing a hypothesis, and (c) shows how this practice is not always sufficient for understanding research results.