bmmtools: validate cognitive measurement models fitted with bmm

bmmtools checks a bmm model before it is used on real data: that it can be estimated from data of a given size, and that its priors imply plausible data. Every check is the same loop, simulate, fit, score.

Simulate

simulate_recovery() draws a data set and the truth that produced it from a bmm model and population values on the link scale. recovery_formula() gives every free parameter a random intercept.

Fit

fit_cached() fits once and reuses the saved fit while nothing that determines it has changed. check_convergence() is the convergence gate. recovery_grid() runs simulate, fit and score over a design grid with replications, one file per cell.

Score

recover() and recover_subjects() score population-level and person-level estimates against the truth; summary() of the result gives bias, RMSE, coverage, interval width and the correlations. extract_correlations() and recover_correlations() do the same for between-subject correlations. prior_check() summarises prior-predictive draws on the observable scale. plot_recovery() and plot_prior_check() draw both.

The articles on the package website walk through each step: https://www.gfrischkorn.org/bmmtools/.

See also

Useful links:

• https://github.com/GidonFrischkorn/bmmtools

• https://www.gfrischkorn.org/bmmtools/

• Report bugs at https://github.com/GidonFrischkorn/bmmtools/issues

Author

Maintainer: Gidon T. Frischkorn gfrischkorn@icloud.com (ORCID) [copyright holder]

Authors:

• Gidon T. Frischkorn gfrischkorn@icloud.com (ORCID) [copyright holder]