Comparing Methods for Estimating Discrepancy in Dyads: A Monte Carlo Simulation

• In Event: 65.060 - Multilevel Modeling: Nonnested Multilevel Data, Measurement Invariance, and Bayesian Approaches

Abstract

The focus of this study is to evaluate methods of estimating dyadic discrepancy, the difference between two individuals. The literature review identified five methods for estimating dyadic discrepancy that are used in current research: raw score difference (RSD), standardized score difference (DIZ), residual score difference (RES), Empirical Bayes estimate of slope in multilevel modeling (EBD), and structural equation modeling estimate (SEM). A Monte Carlo simulation will show which of these five methods generates the most accurate discrepancy estimate. The accuracy of prediction when using the discrepancy to predict an outcome will also be evaluated. The findings will provide methodological guidance to researchers interested in dyadic discrepancies. 

Authors

• Amber McEnturff, University of North Texas

• Qi Chen, University of North Texas

• Wen Luo, Texas A&M University - College Station