R: GLM (and GLMM) family objects for 'asreml'

family_dist {asreml}

R Documentation

GLM (and GLMM) family objects for `asreml`

Description

Family functions specify the details of the models accepted by the family argument to asreml used for fitting generalized linear models (GLMs) and/or generalized linear mixed models (GLMMs).

Usage

asr_gaussian(link = "identity", dispersion = NA)

asr_Gamma(link = "inverse", dispersion = 1, phi = 1)

asr_inverse.gaussian(link = "1/mu^2", dispersion = NA)

asr_binomial(link = "logit", dispersion = 1, total = NULL)

asr_multinomial(link = "logit", dispersion = 1, total = NULL)

asr_negative.binomial(link = "log", dispersion = 1, phi = 1)

asr_poisson(link = "log", dispersion = 1)

Arguments

link

A character string identifying the link function; options and valid names are:

Gaussian: "identity", "log", "inverse" (default = "identity")
Gamma: "identity", "log", "inverse" (default = "inverse")
binomial: "logit", "probit", "cloglog" (default = "logit")
negative.binomial: "identity", "log", "inverse" (default = "log")
poisson: "identity", "log", "sqrt" (default = "log")

dispersion

If NA, the dispersion parameter is estimated, otherwise it is fixed at the nominated value (default = NA for Gaussian and inverse Gaussian models, for the other families default = 1.0).

phi

The known value of the additional parameter phi.

total

A character string or name giving the column in data containing the total counts.

Value

A list of functions and expressions needed by the family argument.

Functions

asr_gaussian(): The Gaussian model/distribution (default).
asr_Gamma(): The gamma model/distribution.
asr_inverse.gaussian(): The inverse Gaussian model/distribution.
asr_binomial(): The binomial model/distribution. If the response is between 0 and 1 it is interpreted as the proportion of successes, otherwise, if not a binary (0,1) variate, it is interpreted as counts of successes; the total number of cases is given by the total argument. If total = NULL, a binary (0,1) response is expected.
asr_multinomial(): The multinomial model/distribution. The response can either be a matrix of counts with the response categories as columns, with an additional column for the total number of cases in each row, or in univariate style with the response as a factor. If the response is a matrix and total = NULL, the total counts are calculated from the category columns.
asr_negative.binomial(): The negative-binomial model/distribution.
asr_poisson(): The poisson model/distribution.