Log-Likelihood of a Fitted Mixture — logLik.shoal

Implementing this gives stats::AIC() and stats::BIC() for free, which is how the number of components is normally chosen.

Usage

# S3 method for class 'shoal_gmm'
logLik(object, ...)

Arguments

object: A fitted shoal_gmm() model.
...: Ignored.

Value

An object of class "logLik", with df and nobs attributes.

Details

The degrees of freedom count the free parameters of a full-covariance mixture: k * p means, k * p * (p + 1) / 2 distinct covariance entries and k - 1 independent mixing proportions.

Examples

fit <- shoal_gmm(as.matrix(iris[, 1:4]), k = 3L)
logLik(fit)
#> 'log Lik.' -180.1957 (df=44)
AIC(fit)
#> [1] 448.3915
BIC(fit)
#> [1] 580.8594