stats (version 3.4.1)

logLik: Extract Log-Likelihood

Description

This function is generic; method functions can be written to handle specific classes of objects. Classes which have methods for this function include: "glm", "lm", "nls" and "Arima". Packages contain methods for other classes, such as "fitdistr", "negbin" and "polr" in package MASS, "multinom" in package nnet and "gls", "gnls" "lme" and others in package nlme.

Usage

logLik(object, …)
# S3 method for lm
logLik(object, REML = FALSE, …)

Arguments

object

any object from which a log-likelihood value, or a contribution to a log-likelihood value, can be extracted.

…

some methods for this generic function require additional arguments.

REML

an optional logical value. If TRUE the restricted log-likelihood is returned, else, if FALSE, the log-likelihood is returned. Defaults to FALSE.

Value

Returns an object of class logLik. This is a number with at least one attribute, "df" (degrees of freedom), giving the number of (estimated) parameters in the model.

There is a simple print method for "logLik" objects.

There may be other attributes depending on the method used: see the appropriate documentation. One that is used by several methods is "nobs", the number of observations used in estimation (after the restrictions if REML = TRUE).

Details

logLik is most commonly used for a model fitted by maximum likelihood, and some uses, e.g.by AIC, assume this. So care is needed where other fit criteria have been used, for example REML (the default for "lme").

For a "glm" fit the family does not have to specify how to calculate the log-likelihood, so this is based on using the family's aic() function to compute the AIC. For the gaussian, Gamma and inverse.gaussian families it assumed that the dispersion of the GLM is estimated and has been counted as a parameter in the AIC value, and for all other families it is assumed that the dispersion is known. Note that this procedure does not give the maximized likelihood for "glm" fits from the Gamma and inverse gaussian families, as the estimate of dispersion used is not the MLE.

For "lm" fits it is assumed that the scale has been estimated (by maximum likelihood or REML), and all the constants in the log-likelihood are included. That method is only applicable to single-response fits.

References

For logLik.lm:

Harville, D.A. (1974). Bayesian inference for variance components using only error contrasts. Biometrika, 61, 383--385.

See Also

logLik.gls, logLik.lme, in package nlme, etc.

AIC

Examples

x <- 1:5
lmx <- lm(x ~ 1)
logLik(lmx) # using print.logLik() method
utils::str(logLik(lmx))

## lm method
(fm1 <- lm(rating ~ ., data = attitude))
logLik(fm1)
logLik(fm1, REML = TRUE)

utils::data(Orthodont, package = "nlme")
fm1 <- lm(distance ~ Sex * age, Orthodont)
logLik(fm1)
logLik(fm1, REML = TRUE)