Martin ModrákMar 19
Preprint alert: Simulation-based validation of Bayes Factor computation with @paul_buerkner and Sebastian Stroppel. We bring lessons learned in improving SBC to validation of Bayes factors. The main idea is still the same: simulate data from the models, fit those and see if the inferences are calibrated. 1/
https://arxiv.org/abs/2508.11814
#Bayesian #stats #rstats #SBC
Show threadRichard D. MoreyMar 20
@modrak_m @avehtari @paul_buerkner This looks very cool!
Show threadRichard D. MoreyMar 20
@modrak_m @avehtari @paul_buerkner I think I caught a bit of shade there for the transparency of the BayesFactor package in the discussion :)
Show threadMartin Modrák
@richarddmorey @avehtari @paul_buerkner The package is still very good though! One thing that made the process a bit challenging was that the posterior draws for coefficients are not the betas themselves but are transformed, but it is not perfectly clear how. So that added another moving part when I tried to make my sims match the package output. But IME this sort of paper vs. package reporting things differently just happens very often with math/stats papers.
Mar 22 at 7:05pmWeb
Show threadRichard D. MoreyMar 23
@modrak_m @avehtari @paul_buerkner Yeah, there are functions that do the transformations but I assumed no one would be interested in the untransformed ones -
Show threadRichard D. MoreyMar 23
@modrak_m @avehtari @paul_buerkner - and that they might even confuse most people, who wouldn’t know what to do with them