Probability of Superiority Effect Size: Performance of Different Estimates

• In Event: Roundtable Session 97
  In Roundtable Session: The Magnitude and Direction of Effect Size Research

Abstract

Probability of superiority (PS) is the probability that a randomly selected observation from population1 is larger than that from population2. PS could be measured by (1) the common language effect size (CL) or (2) PSu. Several formulas that convert popular ES’s to CL are also considered. This simulation study aims at:
1. comparing the performance of CL and PSu with respect to estimation bias and standard error when (1) populations are normal and have equal variance and when (2) normality and homoscedasticity assumptions are violated;
2. examining the sensitivity of conversion formulas for CL to nonnormality and heteroscedasticity.
Results will help researchers choose appropriate estimators of PS under various conditions and encourage the use of robust PS estimators in meta-analysis.

Authors

• Yiran Dong, Indiana University - Bloomington

• Chao-Ying J. Peng, Indiana University