Post

happy to announce the official release of my first R package on CRAN! 🎉 building on the Hmetad toolbox by @smfleming.bsky.social, the hmetad package allows users to fit the meta-d' model of confidence ratings using a familiar brms/lme4-style formula syntax

for anyone unfamiliar with Steve's toolbox, the point of these models is to allow independent estimation of task-level psychological factors (i.e., stimulus sensitivity and response bias) from metacognitive factors (e.g., metacognitive sensitivity and bias) see this paper for reference:

beyond a *ton* of efficiency upgrades 🚀, the package allows for arbitrary hierarchical structure, easy interfacing to other packages in the Stan ecosystem, simple computation of model-implied estimates (e.g., mean confidence, type 1/type 2 ROCs), and a bunch of other cool features

one thing I'm particularly happy about is the ability to plug in arbitrary signal distributions- for example, we allow for metacognitive signal detection with the Gumbel-min distribution (cc @singmann.bsky.social) osf.io/preprints/ps...

another neat feature is that we have implemented a new measure of metacognitive bias that (unlike mean confidence) is independent from task-level performance and metacognitive sensitivity. we have a paper on this measure in progress so stay tuned!