Effect Size Measure for Mediation Analysis with Multicategorical Predictor
Many currently available effect size measures for mediation have limitations when the categorical predictor has three or more categories. Extending the work by Lachowicz, Preacher, and Kelley (2017), I adopted an effect size measure, upsilon υ, for such situation. I conducted a simulation study to investigate the performance of its estimators. Five manipulated factors for data generation (number of groups, sample size per group, Effect size of ai paths, size of b path, Effect size of ci' paths) and one manipulated factor of estimation (R2 shrinkage method) were included in the simulation study. Olkin-Pratt Extended adjusted υ ̃ (υ ̃OPE) showed the best performance across conditions and was recommended for mediation analyses with a multicategorical predictor. Several situations that researchers need to be cautious when using υ ̃OPE were mentioned.
Cao, Zihuan, "Effect Size Measure for Mediation Analysis with Multicategorical Predictor" (2019). ETD Collection for Fordham University. AAI22616839.