loglin.model: define a log-linear model
In hmmm: Hierarchical Multinomial Marginal Models

Description Usage Arguments Details Value Note References See Also Examples

Function to specify a hierarchical log-linear model. This is a particular case of a hmm model.

1 2	loglin.model(lev, int = NULL, strata = 1, dismarg = 0, type = "b", D = TRUE, c.gen = TRUE, printflag = FALSE, names = NULL, formula = NULL)

`lev`	Vector of number of categories of variables
`int`	Generating class of the log-linear model (must be a list) or list of all the interactions included
`strata`	Number of strata
`dismarg`	List of interactions constrained by inequalities - see ‘hmmm.model’
`type`	"b" for baseline logits, "l" for local logits
`D`	Input argument for inequalities - see ‘hmmm.model’
`c.gen`	If FALSE the input int must be the list of the minimal interaction sets to be excluded
`printflag`	If TRUE information on the included and excluded interactions are given
`names`	A character vector whose elements are the names of the variables
`formula`	A formula describing a log-linear model

This function simplifies ‘hmmm.model’ in the case of log-linear models. If formula is employed, c.gen and int must not be declared while names must be specified.

An object of the class hmmmmod defining a log-linear model that can be estimated by ‘hmmm.mlfit’.

If int and formula are not supplied a saturated log-linear model is defined. For log-linear models where the parameters depend on covariates first define a saturated log-linear model and then use the function ‘create.XMAT’.

Agresti A (2012) Categorical data Analysis, (3ed), Wiley, New York.

Bergsma W, Croon M, Hagenaars JA (2009) Marginal Models for Dependent, Clustered, and Longitudinal Categorical Data. Springer.

hmmm.model, hmmm.mlfit, create.XMAT

data(madsen)
y<-getnames(madsen)
names<-c("Infl","Sat","Co","Ho")

f<-~Co*Ho+Sat*Co+Infl*Co+Sat*Ho+Infl*Sat
model<-loglin.model(lev=c(3,3,2,4),formula=f,names=names)

# alternatively 
# model<-loglin.model(lev=c(3,3,2,4),
# int=list(c(3,4),c(2,3),c(1,3),c(2,4),c(1,2)),names=names)

mod<-hmmm.mlfit(y,model,maxit=3000)
print(mod,printflag=TRUE)