pBIC: Pseudo Bayesian Information Criterion

In mcglm: Multivariate Covariance Generalized Linear Models

Pseudo Bayesian Information Criterion

Description

Computes the pseudo Bayesian information criterion (pBIC) for fitted multivariate covariance generalized linear models. The pBIC is defined as

pBIC = -2 \, \ell_p + \text{df} \log(n),

where \ell_p is the pseudo log-likelihood, \text{df} is the effective number of parameters, and n is the total number of observed responses.

This criterion provides a more strongly penalized alternative to pAIC, favoring more parsimonious models when comparing mcglm fits to the same data.

Usage

pBIC(object, verbose = TRUE)

Arguments

object	An object of class mcglm or a list of such objects. When a list is supplied, the pseudo log-likelihood and the number of observations are computed by aggregating information across all models.
verbose	Logical indicating whether the pBIC value should be printed to the console. Defaults to TRUE.

Details

The sample size n used in the penalty term corresponds to the total number of observed responses, obtained from the observed component of the fitted mcglm object(s). As the pBIC is based on a pseudo log-likelihood, it should be used cautiously and only for relative comparisons among models fitted to the same data set.

Value

An (invisible) named list with a single element:

pBIC

A numeric value giving the pseudo Bayesian information criterion associated with the fitted model(s).

Author(s)

Wagner Hugo Bonat, [e-mail]

Source

Bonat, W. H. (2018). Multiple Response Variables Regression Models in R: The mcglm Package. Journal of Statistical Software, 84(4), 1–30.

See Also

gof, plogLik, ESS, pAIC, pKLIC, GOSHO, RJC

mcglm documentation built on Jan. 9, 2026, 1:07 a.m.