An error rate analysis of experimental data testing Nash refinements

Abstract

Error rate analysis applied to experimental data testing choice theories proves useful because it uses the entire distribution of observations - over predicted and unpredicted outcomes - to test theories. In this paper we apply error rate analysis to experimental data testing ash refinements. Evaluating theories using the proportion of consistent responses ignores systematic variation over inconsistent outcomes; a scale of inconsistency is necessary to evaluate refinements as some inconsistent observations are more supportive of a refinement than others. Previous conclusions about the accuracy of refinements reached using the proportion of consistent responses are reversed by the error rate analysis.

Additional details