Skip to contents

The hmetad package is the most recent implementation of the meta-d’ model, developed by Brian Maniscalco and Hakwan Lau (Maniscalco & Lau, 2012). This model has had several implementations since its creation.

The first implementation of the model used maximum likelihood estimation for single participant data (Maniscalco & Lau, 2012), and is still available for download here. The original code is written in MATLAB and a version is also available in Python.

The model was later implemented by (Fleming, 2017) in a hierarchical Bayesian framework, which has been shown to provide much more reliable estimates in the relatively small sample sizes commonly used in psychological experiments. This version, known as the Hmeta-d toolbox, was implemented in the probabilistic programming language JAGS, which in turn has interfaces in both MATLAB and in R.

The hmetad package builds on these previous versions through implementation in the brms package in R, retaining the hierarchical Bayesian approach of the Hmeta-d toolbox while also allowing for flexible estimation of parameters within arbitrarily complex regression designs. Additionally, brms uses the probabilistic programming language Stan, which permits much more efficient sampling with more reliable model convergence warnings and extensive diagnostics. Because of its increased efficiency and flexibility, the hmetad package is our recommended approach to fitting the meta-d’ model.

References

Fleming, S. M. (2017). HMeta-d: Hierarchical bayesian estimation of metacognitive efficiency from confidence ratings. Neuroscience of Consciousness, 2017(1), nix007.
Maniscalco, B., & Lau, H. (2012). A signal detection theoretic approach for estimating metacognitive sensitivity from confidence ratings. Consciousness and Cognition, 21(1), 422–430.