Partial and Semi-Partial Correlations for Categorical Variables in Educational Research:Addressing Two Common Misconceptions

Abstract

This article addresses two misconceptions that educational researchers have about partial and semi-partial correlations. To undo the two misconceptions, this article provides a detailed discussion of the two types of correlations under linear regression which includes multiple dummy indicator variables created for a categorical predictor. A numerical example is provided to demonstrate how the two correlations can be properly utilized to solve research questions in education: 1) Interpretation under the substantive problem and 2) assessment of relative importance of predictors in predicting the dependent variable (DV). The article concludes with a summary of findings and provides the SPSS code used so that interested readers may replicate the results.