gnm (version 0.8-1)

anova.gnm: Analysis of Deviance for Generalized Nonlinear Models

Description

Compute an analysis of deviance table for one or more generalized nonlinear models

Usage

anova.gnm(object, ..., dispersion = NULL, test = NULL)

Arguments

object

an object of class gnm

...

additional objects of class gnm or glm

dispersion

the dispersion parameter for the fitting family. By default it is derived from object

test

(optional) a character string, (partially) matching one of "Chisq", "F", or "Cp". See stat.anova.

Value

• An object of class "anova" inheriting from class "data.frame".

Warning

The comparison between two or more models will only be valid if they are fitted to the same dataset. This may be a problem if there are missing values and R's default of na.action = na.omit is used; an error will be given in this case.

Details

Specifying a single object gives a sequential analysis of deviance table for that fit. The rows of the table show the reduction in the residual deviance and the current residual deviance as each term in the formula is added in turn. If more than one object is specified, the rows of the table show the residual deviance of the current model and the change in the residual deviance from the previous model. (This only makes statistical sense if the models are nested.) It is conventional to list the models from smallest to largest, but this is up to the user. If test is specified, the table will include test statistics and/or p values for the reduction in deviance. For models with known dispersion (e.g., binomial and Poisson fits) the chi-squared test is most appropriate, and for those with dispersion estimated by moments (e.g., 'gaussian', 'quasibinomial' and 'quasipoisson' fits) the F test is most appropriate. Mallows' Cp statistic is the residual deviance plus twice the estimate of $\sigma^2$ times the residual degrees of freedom, which is closely related to AIC (and a multiple of it if the dispersion is known).

See Also

gnm, anova

Examples

set.seed(1)
data(occupationalStatus)

##  Fit a uniform association model separating diagonal effects
Rscore <- scale(as.numeric(row(occupationalStatus)), scale = FALSE)
Cscore <- scale(as.numeric(col(occupationalStatus)), scale = FALSE)
Uniform <- glm(Freq ~ origin + destination + Diag(origin, destination) + 
               Rscore:Cscore, family = poisson, data = occupationalStatus)

##  Fit an association model with homogeneous row-column effects
RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) +
               Nonlin(MultHomog(origin, destination)), family = poisson,
               data = occupationalStatus)

##  Fit an association model with separate row and column effects
RC <- gnm(Freq ~ origin + destination + Diag(origin, destination) +
          Mult(origin, destination), family = poisson,
          data = occupationalStatus)

anova(RC, test = "Chisq")

anova(Uniform, RChomog, RC, test = "Chisq")