optim.const.ordi: performs the M step for the measurement distribution...

In abureau/LCAextend: Latent Class Analysis (LCA) with familial dependence in extended pedigrees

Description

Estimates the cumulative logistic coefficients alpha in the case of multinomial (or ordinal) data with an ordinal constraint on the parameters.

Usage

optim.const.ordi(y, status, weight, param, x = NULL, var.list = NULL)

Arguments

y	a matrix of discrete (or ordinal) measurements (only for symptomatic subjects),
status	symptom status of all individuals,
weight	a matrix of n times K of individual weights, where n is the number of individuals and K is the total number of latent classes in the model,
param	a list of measurement density parameters, here is a list of alpha,
x	a matrix of covariates (optional). Default id NULL,
var.list	a list of integers indicating which covariates (taken from x) are used for a given type of measurement

Details

the constraint on the parameters is that, for a symptom j, the rows alpha[[j]][k,] are equal for all classes k except the first values. Therefore, maximum likelihood estimators are not explicit and the function lrm of the package rms is used to perform a numerical optimization.

Value

The function returns a list of estimated parameters param satisfying the constraint.

Examples

#data
data(ped.ordi)
status <- ped.ordi[,6]
y <- ped.ordi[,7:ncol(ped.ordi)]
data(peel)
#probs and param
data(probs)
data(param.ordi)
#e step
weight <- e.step(ped.ordi,probs,param.ordi,dens.prod.ordi,peel,x=NULL,
                 var.list=NULL,famdep=TRUE)$w
weight <- matrix(weight[,1,1:length(probs$p)],nrow=nrow(ped.ordi),
                 ncol=length(probs$p))
#the function
optim.const.ordi(y[status==2,],status,weight,param.ordi,x=NULL,
                 var.list=NULL)

abureau/LCAextend documentation built on May 3, 2019, 9:41 p.m.