hmmm (version 1.0.0)

hmmm.mlfit: fit a hmm model

Description

Function to estimate a hierarchical multinomial marginal model.

Usage

hmmm.mlfit(y, model, noineq = TRUE, maxit = 1000, 
norm.diff.conv = 1e-05, norm.score.conv = 1e-05,
y.eps = 0, chscore.criterion = 2,
m.initial = y, mup = 1, step = 1)

Arguments

y

A vector of frequencies of the contingency table

model

An object created by `hmmm.model'

noineq

If TRUE inequality constraints specified in the model are ignored

maxit

Maximum number of iteractions

norm.diff.conv

Convergence criterium value on the parameters

norm.score.conv

Convergence criterium value on the constraints

y.eps

Non-negative constant to be added to the original counts in y

chscore.criterion

If equal to zero, convergence information are printed at every iteration

m.initial

Initial estimate of m (expected frequencies)

mup

Weight for the constraints penalty part of the merit function

step

Interval length for the line search

Value

• An object of the class hmmmfit; an estimate of a marginal model defined by `hmmm.model'. The output can be displayed using `summary' or `print', see the help of these functions.

Details

A sequential quadratic procedure is used to maximize the log-likelihood function under inequality and equality constraints. This function calls the procedure `mphineq.fit' which is a generalization of the procedure `mph.fit' by Lang (2004).

References

Bartolucci F, Colombi R, Forcina A (2007) An extended class of marginal link functions for modelling contingency tables by equality and inequality constraints. Statistica Sinica, 17, 691-711. Bergsma WP, Rudas T (2002) Marginal models for categorical data. The Annals of Statistics, 30, 140-159. Lang JB (2004) Multinomial Poisson homogeneous models for contingency tables. The Annals of Statistics, 32, 340-383.

See Also

hmmm.model, hmmm.model.X, summary.hmmmfit, print.hmmmfit

Examples

data(relpol)
y<-getnames(relpol,st=12)
# 1 = Religion, 2 = Politics
names<-c("Rel","Pol")
marglist<-c("l-m","m-g","l-g")
marginals<-marg.list(marglist,mflag="m")

# Hypothesis of stochastic independence: all log odds ratios are null 
model<-hmmm.model(marg=marginals,lev=c(3,7),sel=c(9:20),names=names)
fitmodel<-hmmm.mlfit(y,model)
print(fitmodel, aname="Independence model",printflag=TRUE)
summary(fitmodel)