A Note on Segmented Multinomial Logit Models for the Analysis of Discrete Choice Experiment Data

This note discusses the application of segmented multinomial logit models (MNL) for the analysis of discrete choice experiments. It is shown how the segmentation can be achieved by relying on the model-based recursive partitioning framework and how it can be carried out in R. The approach is illustrated by revisiting an empirical study. The results obtained by segmenting the MNL are compared with the results of the latent class logit model of the original study. Recursively partitioning the MNL leads to comparable but more distinctly interpretable results as the latent class logit model. Potentials and limitations of the approach as well as directions for further research are discussed.