mcIRT (version 0.41)

model_comp: Compare different models

Description

This function makes it easy to compare different models based on different design matrices.

Usage

model_comp(object, ...)

Arguments

object

An object of class nrm or nlm, which is the outcome of fitting a nominal response model nrm() or a nested logit model nelm(). This should be the most parsimonious model, e.g. the model which suspects no DIF.

...

Insert different design matrices here (Input of at least one design-matrix is required!). Use the help function designTemp() to create different designs. These designs will be transformed into a design matrix, and the supplied data will be fitted with these different designs. These competing models are typically using more parameters than the reference model because they often assume some kind of group x item interaction.

Value

A list with 2 elements

object

the submitted object - which is the reference object

modres

a list of fitted objects, which used the same data like the reference object but different designs

Details

The main advantage of this function is first, that different models with different design-matrices are fitted in one step and second that reasonable starting values are used, based on the values of the submitted model object. This should accelerate the model fitting process (by reducing the number of iterations until convergence).

The loop which fits the models always try()s to fit the model, which means that an output is returned even if one or more models fail to converge.

The resulting object can be used to apply likelihood-ratio tests by using the lrt() function.

See Also

designTemp lrt reshMG nrm nelm

Examples

## Not run: 
# # Don't run this example because it takes more than 5sec.
# NUMBI <- 5
# 
# ParList <- lapply(1:NUMBI,function(x)
# {
#   Item1 <- c(c(-2,-1,1,2),c(-1.2,0.3,0.2,0.7))
#   names(Item1) <- c(paste("zeta",1:4,sep=""),paste("lamb",1:4,sep=""))
#   Item1
# })
# 
# 
# names(ParList) <- paste("item",1:NUMBI,sep="")
# 
# set.seed(111)
# 
# perp1 <- rnorm(5000,0,1)
# perp2 <- rnorm(5000,1,1)
# 
# simdat1 <- NRM.sim(ParList,perp1)
# simdat2 <- NRM.sim(ParList,perp2)
# 
# simdat1 <- data.frame(ID=1:5000,simdat1)
# simdat2 <- data.frame(ID=5001:10000,simdat2)
# 
# simdatalla <- merge(simdat1,simdat2,all=TRUE)
# simdatall  <- simdatalla[,-1]
# 
# head(simdatall)
# gruAB <- factor(rep(c("A","B"),each=5000))
# 
# DAT1 <- data.frame(simdatall,ABgroup = gruAB)
# 
# head(DAT1)
# 
# reshOBJ <- reshMG(DAT1,items=1:NUMBI,groups=NUMBI+1,correct=rep(0,NUMBI))
# 
# 
# # ---------------------------------------------------------------------------------
# 
# ergNRM <- nrm(reshOBJ)
# 
# 
# mydes <- designTemp(ngru=2,nit=NUMBI,TYPE="NRM")
# mydes[[1]][2,1] <- 2
# 
# erg_modc <- model_comp(ergNRM,mydes)
# 
# 
# ## End(Not run)