profile.glm {stats}R Documentation

Method for Profiling glm Objects

Description

Investigates the profile log-likelihood function for a fitted model of class "glm".

Usage

## S3 method for class 'glm'
profile(fitted, which = 1:p, alpha = 0.01, maxsteps = 10,
        del = zmax/5, trace = FALSE, test = c("LRT", "Rao"), ...)

Arguments

fitted

the original fitted model object.

which

the original model parameters which should be profiled. This can be a numeric or character vector. By default, all parameters are profiled.

alpha

highest significance level allowed for the profile z-statistics.

maxsteps

maximum number of points to be used for profiling each parameter.

del

suggested change on the scale of the profile t-statistics. Default value chosen to allow profiling at about 10 parameter values.

trace

logical: should the progress of profiling be reported?

test

profile Likelihood Ratio test or Rao Score test.

...

further arguments passed to or from other methods.

Details

The profile z-statistic is defined either as (case test = "LRT") the square root of change in deviance with an appropriate sign, or (case test = "Rao") as the similarly signed square root of the Rao Score test statistic. The latter is defined as the squared gradient of the profile log likelihood divided by the profile Fisher information, but more conveniently calculated via the deviance of a Gaussian GLM fitted to the residuals of the profiled model.

Value

A list of classes "profile.glm" and "profile" with an element for each parameter being profiled. The elements are data-frames with two variables

par.vals

a matrix of parameter values for each fitted model.

tau or z

the profile t or z-statistics (the name depends on whether there is an estimated dispersion parameter.)

Author(s)

Originally, D. M. Bates and W. N. Venables. (For S in 1996.)

See Also

glm, profile, plot.profile

Examples

options(contrasts = c("contr.treatment", "contr.poly"))
ldose <- rep(0:5, 2)
numdead <- c(1, 4, 9, 13, 18, 20, 0, 2, 6, 10, 12, 16)
sex <- factor(rep(c("M", "F"), c(6, 6)))
SF <- cbind(numdead, numalive = 20 - numdead)
budworm.lg <- glm(SF ~ sex*ldose, family = binomial)
pr1 <- profile(budworm.lg)
plot(pr1)
pairs(pr1)

[Package stats version 4.4.0 Index]