pscl (version 1.5.5.1)

predprob.glm: Predicted Probabilities for GLM Fits

Description

Obtains predicted probabilities from a fitted generalized linear model object.

Usage

# S3 method for glm
predprob(obj, newdata = NULL, at = NULL, ...)

Value

A matrix of predicted probabilities. Each row in the matrix is a vector of probabilities, assigning predicted probabilities over the range of responses actually observed in the data. For instance, for models with family=binomial, the matrix has two columns for the "zero" (or failure) and "one" (success) outcomes, respectively, and trivially, each row in the matrix sums to 1.0. For counts fit with family=poisson or via glm.nb, the matrix has length(0:max(y)) columns. Each observation used in fitting the model generates a row to the returned matrix; alternatively, if

newdata is supplied, the returned matrix will have as many rows as in newdata.

Arguments

obj

a fitted object of class inheriting from "glm"

newdata

optionally, a data frame in which to look for variables with which to predict. If omitted, the fitted linear predictors are used.

at

an optional numeric vector at which the probabilities are evaluated. By default 0:max(y) where y is the original observed response.

...

arguments passed to or from other methods

Author

Simon Jackman

Details

This method is only defined for glm objects with family=binomial or family=poisson, or negative binomial count models fit with the glm.nb function in library(MASS).

See Also

predict.glm

Examples

data(bioChemists)
glm1 <- glm(art ~ .,
            data=bioChemists,
            family=poisson,
            trace=TRUE)  ## poisson GLM
phat <- predprob(glm1)
apply(phat,1,sum)                    ## almost all 1.0