Does the measure of dispersion matter in multilevel research? A comparison of the relative performance of dispersion indexes

Within the context of climate strength, this simulation study examines the validity of various dispersion indexes for detecting meaningful relationships between variability in group member perceptions and outcome variables. We used the simulation to model both individual-and group-level phenomena, vary appropriate population characteristics, and test the proclivity of standard and average deviation, interrater agreement indexes (r(wg) r*(wg), a(wg)), and coefficient of variation (both normed and unnormed) for Type I and Type II errors. The results show that the coefficient of variation was less likely to detect interaction effects although it outperformed other measures when detecting level effects. Standard deviation was shown to be inferior to other indexes when no level effect is present although it may be an effective measure of dispersion when modeling strength or interaction effects. The implications for future research, in which dispersion is a critical component of the theoretical model, are discussed.