A Bayesian hierarchical model of lexicographic choice

Abstract

I present a lexicographic, threshold-based model of choice used to evaluate decision makers' preferences among risky alternatives. Using a hierarchical Bayesian frame-work, this model is able to account for observed individual differences by allowing for variable threshold values in attribute features, as well as the order that individuals consider attributes of the choice alternatives. Performance of the model is evaluated via a parameter recovery test using simulated data. I also apply the model to the choice data from a decision-making-under-risk experiment (Davis-Stober, Brown and Cavagnaro, 2015). Bayesian p-values are obtained to check the model fits for every individual, and sensitivity analysis is carried out to measure the degree to which choices of prior distributions affect the results. Finally, I discuss the implications of the Bayesian hierarchical model of lexicographic choice I present in this paper.