|
Abstract:
|
Delay discounting is a behavioral economic concept that measures the subjective rate at which an individual devalues a future reward as a function of the delay to receiving the reward. These devaluation rates are known to correlate with a host of problematic behaviors including excessive gambling, overeating, drug use, and alcohol abuse. While many models have been proposed to characterize delay discounting, there is little consensus among practitioners about which models are best. Difficulty is compounded since practitioners are frequently reluctant to cede model choice to statistical approaches that are perceived to be naïve to domain specific knowledge. Since many of the top nonlinear models are nested, there is an opportunity to stylize Bayesian model selection using mixture models as priors for specific parameters. In this approach, practitioners' model preferences are encoded as prior beliefs that are incorporated into the selection procedure. In this talk I will demonstrate a fully Bayesian approach to model selection. Emphasis will be on selection, parameter estimation, and the impact of prior model belief on resulting inference for a variety of delay discounting data.
|
