What it’s about

In this paper, we present a Bayesian estimation workflow for generalized log-linear Poisson item-response models applied to fluency scores in divergent-thinking assessment, with practical implementation using brms and Stan.

Abstract

Divergent-thinking tests often rely on fluency scores (counts of ideas generated per prompt), and these counts can be modeled with Poisson item-response models. This paper revisits the Rasch Poisson Counts Model and its generalized two-parameter extension (2PPCM), and shows how both can be estimated in a Bayesian multilevel regression framework through brms (Stan backend). Using an example dataset with three tasks and 202 participants, the article details model specification, estimation, convergence and fit checks, model comparison, and interpretation. It also provides practical guidance for plotting item response functions and deriving overdispersion, reliability, and factor scores.

How to access the paper

Go to the journal page for this article here.

Download the PDF here.