The Speed-Accuracy Trade-Off in a Marginal Bayesian Modelling

Abstract

Current approaches in psychometrics to integrate responses and response times in a single model are limited by assuming a fixed correlation between a test-taker’s ability and working speed across the whole test. In the proposed marginal joint-model, the relationship between the person parameters within a group of test-takers is explicatively modelled for each item of the test. Hypotheses about between-item differences in said relationship and about betweengroup differences in ability or speed are evaluated with approximated Bayes factors. The results of two simulation studies show, a) a satisfying performance in recovering model parameters and b) a sufficient statistical power inBayesian hypothesis testing. The insights gained through the proposed model can be used to create fairer tests for diverse populations.