Predict choice: A comparison of 21 mathematical models

Eric SchulzUniversity College London
Maarten SpeekenbrinkUniversity College London
David R. ShanksUniversity College London

Abstract

How should we choose a model that predicts human choices? Two important factors in this choice are a model's predictive power and a model's fexibility. In this paper, we compare these aspects of models in a large set of models applied to an experiment in which participants chose between brands of fictitious chocolate bars and a quasi-experiment predicting movies' gross revenue. We show that there is a trade-o ff between flexibility and predictive power, but that this trade-o ff appears to lie more towards the "flexible" side than what was previously thought.

Files

Predict choice: A comparison of 21 mathematical models (217 KB)



Back to Table of Contents