Open AccessHierarchies improve individual assessment of temporal discounting behavior.
Author(s) -
M. Fiona Molloy,
Ricardo J. Romeu,
Peter D. Kvam,
Peter R. Finn,
Jerome R. Busemeyer,
B. E. Turner
Publication year - 2020
Publication title -
decision
Language(s) - English
Resource type - Journals
eISSN - 2325-9973
pISSN - 2325-9965
DOI - 10.1037/dec0000121
Subject(s) - impulsivity , discounting , temporal discounting , estimation , bayesian probability , delay discounting , set (abstract data type) , addiction , task (project management) , computer science , psychology , range (aeronautics) , time preference , econometrics , cognitive psychology , machine learning , artificial intelligence , mathematics , clinical psychology , economics , psychiatry , materials science , neoclassical economics , management , finance , composite material , programming language
Delay discounting behavior has proven useful in assessing impulsivity across a wide range of populations. As such, accurate estimation of the shape of each individual's temporal discounting profile is paramount when drawing conclusions about how impulsivity relates to clinical and health outcomes such as gambling, addiction, and obesity. Here, we identify an estimation problem with current methods of assessing temporal discounting behavior, and propose a simple solution. First, through a simulation study we identify types of temporal discounting profiles that cannot reliably be estimated. Second, we show how imposing constraints through hierarchical modeling ameliorates these recovery problems. Finally, we apply our solution to a large data set from a temporal discounting task, and illustrate the importance of reliable estimation within patient populations. We conclude with a brief discussion on how hierarchical Bayesian methods can aid in model estimation, compensate for small samples, and improve predictions of externalizing psychopathology.