Design and analysis of discrete choice experiments with partial profiles in the presence of interactions



In a discrete choice experiment, respondents sometimes make their choices on the basis of only one dominant attribute rather than making trade-offs among all the attributes. For example, in studies involving price as an attribute, respondents may always choose the product profile with the lowest price. Also, a choice task including many attributes may encourage respondent decisions that are not fully compensatory. To thwart these behaviors, the investigator can hold the levels of some of the attributes constant in every choice set. The resulting designs are called partial profile designs. In this talk, we construct D-optimal partial profile designs for estimating interaction-effects models using a Bayesian design algorithm. To determine the constant attributes in each choice set, we provide a generalization of an approach that makes use of balanced incomplete block designs. We present results from an actual experiment in software development combined with a discussion on how to analyze the data on an individual level using a penalized maximum likelihood approach.
This research seminar is organised by the Erasmus Centre for Marketing of Innovation (ECMI).
Contact information:
Dr. G. Liberali