CDM (version 4.8-0)

predict: Expected Values and Predicted Probabilities from Item Response Response Models

Description

This function computes expected values for each person and each item based on the individual posterior distribution. The output of this function can be the basis of creating item and person fit statistics.

Usage

IRT.predict(object, dat, group=1)

## S3 method for class 'din':
predict(object, group=1, \dots)

## S3 method for class 'gdina':
predict(object, group=1, \dots)

## S3 method for class 'mcdina':
predict(object, group=1, \dots)

## S3 method for class 'gdm':
predict(object, group=1, \dots)

## S3 method for class 'slca':
predict(object, group=1, \dots)

Arguments

object

Object for the S3 methods IRT.irfprob and IRT.posterior are defined. In the CDM packages, these are the objects of class din

dat

Dataset with item responses

group

Group index for use

...

Further arguments to be passed.

Value

• A list with following entries
• expected
  Array with expected values (persons $\times$ classes $\times$ items)
• probs.categ
  Array with expected probabilities for each category (persons $\times$ categories $\times$ classes $\times$ items)
• variance
  Array with variance in predicted values for each person and each item.
• residuals
  Array with residuals for each person and each item
• stand.resid
  Array with standardized residuals for each person and each item

Examples

#############################################################################
# EXAMPLE 1: Fitted Rasch model in TAM package
#############################################################################

library(TAM)
data(sim.rasch, package="TAM")

#--- Model 1: Rasch model
mod1 <- tam.mml(resp=sim.rasch) 
# apply IRT.predict function
prmod1 <- IRT.predict(mod1 , mod1$resp )
str(prmod1)

#############################################################################
# EXAMPLE 2: Predict function for din
#############################################################################

data(sim.dina)
data(sim.qmatrix)

# DINA Model
mod1 <- din(sim.dina, q.matr = sim.qmatrix, rule = "DINA" )
summary(mod1)
# apply predict method
prmod1 <- IRT.predict( mod1 , sim.dina )
str(prmod1)