Improving interpretability: gamma as an alternative to R(2) as a measure of effect size.

Journal article

A traditional measure of effect size associated with tests for difference between two groups is the variance explained by group membership (R(2)). If exposure to a disease causes a small but long term deficit in performance, however, R(2) does not capture that cumulating effect. We propose an alternative statistic, gamma, based on the probability of an unexposed person outperforming an exposed person. Although gamma is also a point estimate, it more easily conveys what the cumulating effect of a deficit would be. We discuss some of the advantages of this measure.

Authors: Reidpath, D; Diamond, MR; Hartel, G; Glasziou, P

• Journal: Statistics in medicine
• Volume: 19
• Issue: 10
• Pages: 1295-1302
• Publication date: 2000-05-01
• DOI: 10.1002/(sici)1097-0258(20000530)19:10<1295::aid-sim493>3.0.co;2-z
• EISSN: 1097-0258
• ISSN: 0277-6715
• Language: English
• Keywords: Data Interpretation, Statistical; Monte Carlo Method; Cognition; Otitis Media with Effusion; Language Tests; Child; Humans; Computer Simulation