Compute the corrected marginal AIC (AICc)

Calculates the corrected Akaike Information Criterion \(\mathrm{AICc} = p k + 2 \ell + 2k(k+1)/(n-k-1)\), where \(k\) is the number of estimated parameters, \(\ell\) is the negative log-likelihood at the optimum, \(p\) is the penalty multiplier, and \(n\) is the sample size. When \(n = \infty\) the small-sample correction term vanishes and the result reduces to standard marginal AIC. Compatible with output from both optim ($value) and nlminb ($objective).

Usage

marg_AIC(opt, p = 2, n = Inf)

Arguments

opt

Named list of optimiser output. Must contain $par and either $objective (e.g., nlminb) or $value (e.g., optim).

p

Numeric. Penalty multiplier per parameter. Default 2 (standard AIC).

n

Numeric. Sample size used for the small-sample correction. Default Inf (no correction, equivalent to AIC).

Value

Numeric. The AICc value.