Learn R Programming

aod (version 1.3.3)

AIC-methods: Akaike Information Criteria

Description

Extracts the Akaike information criterion (AIC) and the corrected AIC (AICc) from fitted models of formal class “glimML” and possibly computes derived statistics.

Usage

# S4 method for glimML
AIC(object, ..., k = 2)

Arguments

object

fitted model of formal class “glimML” (functions betabin or negbin).

...

optional list of fitted models separated by commas.

k

numeric scalar, with a default value set to 2, thus providing the regular AIC.

Methods

glimML

Extracts the AIC and AICc from models of formal class “glimML”, fitted by functions betabin and negbin.

Details

\(AIC = -2~\mbox{log-likelihood} + 2*n_{par}\), where \(n_{par}\) represents the number of parameters in the fitted model.
\(AICc = AIC + 2 * n_{par} * (n_{par} + 1) / (n_{obs} - n_{par} + 1)\), where \(n_{obs}\) is the number of observations used to compute the log-likelihood. It should be used when the number of fitted parameters is large compared to sample size, i.e., when \(n_{obs} / n_{par} < 40\) (Hurvich and Tsai, 1995).

References

Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference: a practical information-theoretic approach. New-York, Springer-Verlag, 496 p.
Hurvich, C.M., Tsai, C.-L., 1995. Model selection for extended quasi-likelihood models in small samples. Biometrics, 51 (3): 1077-1084.

See Also

Examples in betabin and see AIC in package stats.