Thinking Critically About When to Convert Odds Ratios Into Mean Differences in Meta-Analysis

• In Event: Understanding Learning Strategies and Student Performance Using Meta-Analytic Methods

Abstract

Literature provides conversion formulas for how to transform odds ratios (effect size measures of dichotomous outcomes) for combination with continuous outcomes’ standardized mean differences, but minimal guidance on under what circumstances to convert. Most meta-analyses do not delineate how dichotomous outcomes are included/rescaled. In this method note, we propose augmenting existing meta-analysis processes by incorporating explicit consideration of when/whether to convert. We provide an empirical example when merging might be inadvisable, propose initial guidelines for when to transform ORs, and suggest next questions.

Authors

• Dan Fitzpatrick, University of Michigan

• James Eric Pustejovsky, University of Wisconsin - Madison

• Joshua R. Polanin, American Institutes for Research

• Terri D. Pigott, Georgia State University