Non-Parametric Preference Estimation

For general/non-budgetary datasets, Prest v0.9.11 can estimate non-parametrically which model (or models) in its toolkit is the best match for a given subject in the dataset by identifying “how far” each model is from fully explaining that person’s choices.

Generalizing the Houtman-Maks [HM85] method in the model-based way that was suggested in [CGCGT16], Prest v0.9.11 computes the distance score associated with every user-selected model for every subject in the dataset. This corresponds to the number of observations that need to be removed from a subject’s data in order for the remaining choices to be fully compatible with the model in question. Prest also provides information about the compatible instances of every model that is optimal in this sense.

These model- and preference-estimation features enable users to analyse the available data to test for the proximity of choice behavior not only with utility maximization but also with several more general models that provide explanations of well-documented behavioral phenomena such as context-dependent choice, cyclic choice, status quo bias, choice deferral and choice overload.