bayestestR 0.18.0

New functionality

• Improved Bayes factor methods:

  • New docs at ?bayesfactor_methods

  • as.matrix() for bayesfactor_restricted(), to obtain a matrix of Bayes factors between all restricted models.

• Added support for CmdStanFit models from {cmdstanr} and expanded support for stanfit models from rstan.

Changes

• as.matrix(<bf>) now returns class bayesfactor_matrix and has a simpler printing.

• diagnostic_posterior() works with 'raw' MCMC samples (i.e., lists of data
  frames or matrices representing samples of parameters from chains, or 3D arrays)
  as well as objects from rstanarm/brms/lavaan models.

• diagnostic_posterior() now reports the tail-ESS (the minimum of the
  effective sample sizes for the 5% and 95% quantiles) in the ESS column,
  instead of the basic n_eff from older Stan versions. The tail-ESS is more
  relevant for assessing the reliability of credible intervals and other
  tail-based quantities. To also obtain the bulk-ESS (useful for central
  tendency estimates), pass "ESS_bulk" to the diagnostic argument.

• effective_sample() for stanfit objects now also returns the tail-ESS
  (ESS_tail), consistent with brmsfit and stanreg objects.