The Effect of Scale Changes on Meta-Analysis of Multiplicative and Main Effects Models

Abstract

It is well-known that when researchers explore conditional relationships, they have to be quite careful about the impact of scaling on the size of these relationships. In this article, the impact of such changes on meta-analyses of conditional relationships is explored. A number of parameters (correlations, regression coefficients, standardized regression coefficients, semipartial correlations, and partial correlations) are examined. Their strengths and weaknesses are assessed. It is concluded that the most nearly appropriate parameters for researchers to aggregate are unstandardized regression coefficients or partial correlation coefficients.