QTEST: An Order-constrained Inference Framework for Testing Theories of Pairwise Choice

Abstract

Many nonparametric probabilistic choice models for binary paired comparisons are characterized by order-constrained parameter spaces. The talk gives an overview of distribution-free random utility models, distance-based probabilistic specifications, and nonparametric Fechnerian models. Each of these models features a parameter space consisting of a disjoint union of convex polytopes. Hence, frequentist hypothesis testing and Bayesian model selection depend on suitable order-constrained methods. The talk will present a public domain software package, QTEST, designed to carry out such analyses.