< Back to previous page

Project

Discrete choice experiments: design software and two-stage choice models

Discrete choice experiments (DCE's) are commonly used to elicit preferences from decision makers. This thesis contributes to designing such experiments, as well as modelling the choice data gathered by such experiments.

In the second chapter, the newly developed R-package "idefix" is presented. The latter enables users to generate optimal designs for DCE’s. Furthermore, it includes adaptive design methodology and the option to gather empirical choice data by setting up surveys. The main algorithms optimize designs assuming a conditional or mixed logit model, which are the most frequently applied choice models. As such, the package aims to provide essential tools for conducting DCE’s.

In chapters three and four, two-stage choice models were developed that include a non-compensatory consideration stage and a compensatory choice stage. In chapter three, the SCL model was presented which allows to estimate preference coefficients separately from consideration coefficients. We show that, if the true choice mechanism is one in which the importance of attributes differs across stages, the produced choice data can be captured by the SCL model. The model was compared with benchmark models on empirical choice data concerning student rooms. It was concluded that the SCL model allows to gain a deeper understanding of the consideration process, which was not possible with the other models under review.

In the fourth chapter, a two-stage choice model that accounts for consideration heterogeneity was presented. The LCCL model allows respondents to have different thresholds for accepting attribute levels by making use of latent classes. To model the consideration process, a conjunctive rule was used as an approximation of a variety of other simplifying heuristics. We compared our approach with the SCL, the CL and the mixed logit model (MIXL) on empirical choice data concerning cinemas. The results show that the data is best explained by the proposed model, and suggests that most participants use simplifying heuristics before choosing.

Date:22 Sep 2015 →  27 Aug 2021
Keywords:Discrete choice experiments, Choice modelling
Disciplines:Applied economics, Economic history, Macroeconomics and monetary economics, Microeconomics, Tourism
Project type:PhD project