RaschPLE {plRasch}R Documentation

Fit Rasch Family Models using Pseudolikelihood Esitmation

Description

This function fits Rasch family models using pseudolikelihood esitmation. It is capable of dealing with polytomous items, and multidimensional latent variables.

Usage

RaschPLE(data, item.mtx, trait.mtx)

Arguments

data is a data frame or matrix with rows indicating individuals, columns indicating items, and the entry values indicating the choices.
item.mtx is the adjacency matrix between items and latent traits
trait.mtx is the adjacency matrix for latent traits

Details

The model is

                       exp( w[i,h]' theta[v] + beta[i,h] )
   P(X[v,i] = h) = ---------------------------------------------
                    sum_l exp( w[i,l]' theta[v] + beta[i,l] ) 

where

X[v,i] is the response of vth individual to ith item; w[i,h] is a vector of known category weights or scores for response h of ith item; theta[v] is a vector of latent traits for vth individual; beta[i,h] is the item difficulty parameter for ith item; associated with response h.

The function only returns the item parameter beta.

Essentially, it is a wrapper function: the equvialent llla model is fitted.

Value

coefficients estimated item parameter beta
se standard error of beta
covb covariance matrix of the estimated parameter beta

Author(s)

Zhushan "Mandy" Li & Feng Hong

References

Anderson, C.J., Li, Z., & Vermunt, J.K. (2007). Estimation of models in the Rasch family for polytomous items and multiple latent variables. Journal of Statistical Software, 20.

See Also

llla

Examples

NCAT <- 2;
NITEM <- 4;
NEXAMINEE <- 50;
BETA <- c(-1, 0, 0.5, 1)
set.seed(1);
rasch.sim <- simRasch(ncat=NCAT, nitem=NITEM, nexaminee=NEXAMINEE, beta=BETA)
sim.data <- rasch.sim$data
colnames(sim.data) <- paste("I", 1:NITEM, sep='')
## The model item adjacency matrix and the latent trait adjacency matrix
item.mtx <- rep(1, NITEM);
trait.mtx <- 1;

plfit.rasch <- RaschPLE(sim.data, item.mtx, trait.mtx)
plfit.rasch 

[Package plRasch version 0.1 Index]