Skip to contents

The fitted log-likelihood, with df (parameter count) and nobs attributes attached so base AIC()/BIC() work directly off it.

Usage

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

Arguments

object

a corehydro_fit.

...

unused; present for generic consistency.

Value

An object of class logLik.

Details

fit_gmm() is method-of-moments: GeneralizedMethodOfMoments computes no likelihood surface, so a GMM fit reports NA for $log_likelihood, $aic and $bic, and logLik() returns NA_real_ (still carrying df/nobs) so base AIC()/BIC() come back a self-explanatory NA. The GMM analogue of a likelihood-based goodness-of-fit summary is the J-statistic overidentification diagnostic already on the fit (fit$j_stat/fit$j_stat_pval; see fit_gmm()), which is also what print() on a corehydro_fit shows in place of the log-likelihood line for a GMM fit.