14 Random-Effects Models for Preference Data

Abstract

Publisher Summary Most statistical models for the analysis of choice data also referred to as “preference models,” have been developed for response settings to serve the dual purpose of summarizing choice outcomes and of facilitating the forecasting of choices made by the judges facing possibly new or different variants of the choice options. Thus, preference models can provide useful information to identify both option characteristics that influence choices and the systematic sources of individual differences in the evaluation of these option characteristics. Because preference models focus mainly on choice outcomes, they render little information about the underlying choice processes and the effect of unobserved constraints that frequently are a part of revealed choices. Despite these limitations, preference models can be useful in characterizing a choice process in two important respects, provided multiple choices are observed from the same decision maker. First, they can provide benchmarks for understanding whether decision makers are consistent in making their choices. Second, preference models allow assessing whether decision makers act as utility maximizers.