A tutorial for estimating Bayesian hierarchical mixture models for visual working memory tasks: Introducing the...

Mixture models for visual working memory tasks using continuous report recall are highly popular measurement models in visual working memory research. Yet, efficient and easy-to-implement hierarchical Bayesian estimation procedures that flexibly enable group or condition comparisons are scarce. Specifically, most software packages implementing mixture models have used maximum likelihood estimation for single-subject data. Such estimation procedures require a large number of trials per participant to obtain robust and reliable estimates. This problem can be solved with hierarchical Bayesian estimation procedures that provide robust and reliable estimates with lower trial numbers. In this tutorial, we illustrate how mixture models for visual working memory tasks can be specified and fit in the newly developed R package bmm. The benefit of this implementation over existing hierarchical Bayesian implementations is that bmm integrates hierarchical Bayesian estimation of the mixture models with an efficient linear model syntax that enables us to adapt the mixture model to practically any experimental design. Specifically, this implementation allows for varying model parameters over arbitrary groups or experimental conditions. Additionally, the hierarchical structure and the specification of informed priors can frequently improve subject-level parameter estimation and solve estimation problems. We illustrate these benefits in different examples and provide R code for easy adaptation to other use cases.