For general/non-budgetary datasets, Prest can estimate non-parametrically which model(s) 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 and Maks  method in the model-based way that was first suggested in Costa-Gomes, Cueva, Gerasimou, and Tejiščák , Prest computes the distance score associated with every user-selected model for every subject in the dataset.
This score corresponds to the number of observations that need to be removed from a subject’s data in order for the remaining observations 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 allow you to analyse the available data to test for the proximity of choice behaviour not only with utility maximization but also with several models of general choice that explain well-documented behavioural phenomena such as context-dependent choices, cyclic choices, status-quo biased choices, choice deferral and choice overload.
Prest’s core program is designed to utilise all of your computer’s CPU power by simultaneously engaging all its cores, by default. You can change that by checking the “Disable parallelism” box at the bottom of the “Model estimation” window.
If your dataset includes observations where the deferral/outside option was chosen and you wish to ignore these observations, you can do so by checking the “Disregard deferrals” box at the bottom of the “Model estimation” window.
- Forced-Choice Models (no outside option)
- Non-Forced-Choice Models (feasible outside option)
- Choice Models with a Default/Status-Quo Alternative