Hierarchical Bayesian estimation and hypothesis testing for delay discounting tasks

I am happy to announce my 3rd paper of the year, accepted for publication in Behavior Research Methods. Following my initial foray into writing review papers (2 earlier this year), this is my first methods paper, and also my first contribution to higher-level decision making.

Much of my previous work has examined Bayesian decision making approaches applied to ‘low-level’ perceptual or attentional phenomena, but the present paper examines how we trade off the immediacy v.s. the magnitude of a reward in the temporal discounting task. This is also known as the inter-temporal choice task, or colloquially as the Marshmallow task following the classic experiment by Walter Mischel. This can of course can be carried out with any form of reward (or punishment) such as money, juice, or chocolate.

The paper provides a method (along with Matlab code) to analyse data from delay discounting experiments. I’ve attempted to do this ‘properly’ in that it uses both Bayesian parameter estimation and Bayesian hypothesis testing, and takes a hierarchical approach.

Go get the paper and code here:

Vincent, B., T. (in press) Hierarchical Bayesian estimation and hypothesis testing for delay discounting tasks, Behavior Research Methods. doi:10.3758/s13428-015-0672-2 [pdf | Matlab code on GitHub]

And of course, I did all this so that people actually use these methods to analyse their data. Therefore I’ve made the analysis code available on GitHub, and over the next little while I will be updating the instructions to make it super easy for people to analyse their data in the way I describe in the paper.

While I am not leaving behind my interests in attention and perception, this represents a new focus for me. I have another paper planned on a similar vein. I am also actively looking to collaborate with empiricists on delay discounting topics; please do get in touch.

Go to the Bayesian analysis toolbox for delay discounting data project page for more.

Leave a Reply