Title: Individual model selection
Speaker: Prince Osei, Carleton University
Location: HP 4351 (MacPhail Room) – Carleton University
Date: Friday, March 24, 2023
Time: 1:00 – 2:00 pm
Abstract: In the study of decision-making patterns by individuals, such as the Iowa gambling task (IGT) process, there are several competing models suitable for describing a participant or group of individuals. Selecting a single best overall model typically comes at the cost of misallocating some individuals for the sake of the population. The calibration of the models requires re-running these models several times whenever an individual or set of individuals are moved or reshuffled among these models. Usually, there is a fixed computational budget, and can only afford to run each model once. An individualized model selection strategy that uses the Bayesian information criterion (BIC) is proposed. The method is computationally efficient and easy to implement to allocate the subjects to these competing models. The proposed individualized model selection strategy is illustrated using a hand-picked sample from the IGT data set.