tam2mirt: Converting a fitted 'TAM' Object into a 'mirt' Object

In sirt: Supplementary Item Response Theory Models

Converting a fitted TAM Object into a mirt Object

Description

Converts a fitted TAM object into a mirt object. As a by-product, lavaan syntax is generated which can be used with lavaan2mirt for re-estimating the model in the mirt package. Up to now, only single group models are supported. There must not exist background covariates (no latent regression models!).

Usage

tam2mirt(tamobj)

Arguments

tamobj

Object of class TAM::tam.mml

Value

A list with following entries

mirt	Object generated by mirt function if est.mirt=TRUE
mirt.model	Generated mirt model
mirt.syntax	Generated mirt syntax
mirt.pars	Generated parameter specifications in mirt
lavaan.model	Used lavaan model transformed by lavaanify function
dat	Used dataset. If necessary, only items used in the model are included in the dataset.
lavaan.syntax.fixed	Generated lavaan syntax with fixed parameter estimates.
lavaan.syntax.freed	Generated lavaan syntax with freed parameters for estimation.

See Also

See mirt.wrapper for convenience wrapper functions for mirt objects.

See lavaan2mirt for converting lavaan syntax to mirt syntax.

Examples

## Not run: 
library(TAM)
library(mirt)

#############################################################################
# EXAMPLE 1: Estimations in TAM for data.read dataset
#############################################################################

data(data.read)
dat <- data.read

#**************************************
#*** Model 1: Rasch model
#**************************************

# estimation in TAM package
mod <- TAM::tam.mml( dat )
summary(mod)
# conversion to mirt
res <- sirt::tam2mirt(mod)
# generated lavaan syntax
cat(res$lavaan.syntax.fixed)
cat(res$lavaan.syntax.freed)
# extract object of class mirt
mres <- res$mirt
# print and parameter values
print(mres)
mirt::mod2values(mres)
# model fit
mirt::M2(mres)
# residual statistics
mirt::residuals(mres, type="Q3")
mirt::residuals(mres, type="LD")
# item fit
mirt::itemfit(mres)
# person fit
mirt::personfit(mres)
# compute several types of factor scores (quite slow)
f1 <- mirt::fscores(mres, method='WLE',response.pattern=dat[1:10,])
     # method=MAP and EAP also possible
# item plot
mirt::itemplot(mres,"A3")    # item A3
mirt::itemplot(mres,4)       # fourth item
# some more plots
plot(mres,type="info")
plot(mres,type="score")
plot(mres,type="trace")
# compare estimates with estimated Rasch model in mirt
mres1 <- mirt::mirt(dat,1,"Rasch" )
print(mres1)
mirt.wrapper.coef(mres1)

#**************************************
#*** Model 2: 2PL model
#**************************************

# estimation in TAM
mod <- TAM::tam.mml.2pl( dat )
summary(mod)
# conversion to mirt
res <- sirt::tam2mirt(mod)
mres <- res$mirt
# lavaan syntax
cat(res$lavaan.syntax.fixed)
cat(res$lavaan.syntax.freed)
# parameter estimates
print(mres)
mod2values(mres)
mres@nest   # number of estimated parameters
# some plots
plot(mres,type="info")
plot(mres,type="score")
plot(mres,type="trace")
# model fit
mirt::M2(mres)
# residual statistics
mirt::residuals(mres, type="Q3")
mirt::residuals(mres, type="LD")
# item fit
mirt::itemfit(mres)

#**************************************
#*** Model 3: 3-dimensional Rasch model
#**************************************

# define Q-matrix
Q <- matrix( 0, nrow=12, ncol=3 )
Q[ cbind(1:12, rep(1:3,each=4) ) ] <- 1
rownames(Q) <- colnames(dat)
colnames(Q) <- c("A","B","C")
# estimation in TAM
mod <- TAM::tam.mml( resp=dat, Q=Q, control=list(snodes=1000,maxiter=30) )
summary(mod)
# mirt conversion
res <- sirt::tam2mirt(mod)
mres <- res$mirt
# mirt syntax
cat(res$mirt.syntax)
  ##   Dim01=1,2,3,4
  ##   Dim02=5,6,7,8
  ##   Dim03=9,10,11,12
  ##   COV=Dim01*Dim01,Dim02*Dim02,Dim03*Dim03,Dim01*Dim02,Dim01*Dim03,Dim02*Dim03
  ##   MEAN=Dim01,Dim02,Dim03
# lavaan syntax
cat(res$lavaan.syntax.freed)
  ##   Dim01=~ 1*A1+1*A2+1*A3+1*A4
  ##   Dim02=~ 1*B1+1*B2+1*B3+1*B4
  ##   Dim03=~ 1*C1+1*C2+1*C3+1*C4
  ##   A1 | t1_1*t1
  ##   A2 | t1_2*t1
  ##   A3 | t1_3*t1
  ##   A4 | t1_4*t1
  ##   B1 | t1_5*t1
  ##   B2 | t1_6*t1
  ##   B3 | t1_7*t1
  ##   B4 | t1_8*t1
  ##   C1 | t1_9*t1
  ##   C2 | t1_10*t1
  ##   C3 | t1_11*t1
  ##   C4 | t1_12*t1
  ##   Dim01 ~ 0*1
  ##   Dim02 ~ 0*1
  ##   Dim03 ~ 0*1
  ##   Dim01 ~~ Cov_11*Dim01
  ##   Dim02 ~~ Cov_22*Dim02
  ##   Dim03 ~~ Cov_33*Dim03
  ##   Dim01 ~~ Cov_12*Dim02
  ##   Dim01 ~~ Cov_13*Dim03
  ##   Dim02 ~~ Cov_23*Dim03
# model fit
mirt::M2(mres)
# residual statistics
residuals(mres,type="LD")
# item fit
mirt::itemfit(mres)

#**************************************
#*** Model 4: 3-dimensional 2PL model
#**************************************

# estimation in TAM
mod <- TAM::tam.mml.2pl( resp=dat, Q=Q, control=list(snodes=1000,maxiter=30) )
summary(mod)
# mirt conversion
res <- sirt::tam2mirt(mod)
mres <- res$mirt
# generated lavaan syntax
cat(res$lavaan.syntax.fixed)
cat(res$lavaan.syntax.freed)
# write lavaan syntax on disk
  sink( "mod4_lav_freed.txt", split=TRUE )
cat(res$lavaan.syntax.freed)
  sink()
# some statistics from mirt
print(mres)
summary(mres)
mirt::M2(mres)
mirt::residuals(mres)
mirt::itemfit(mres)

# estimate mirt model by using the generated lavaan syntax with freed parameters
res2 <- sirt::lavaan2mirt( dat, res$lavaan.syntax.freed,
            technical=list(NCYCLES=3), verbose=TRUE)
                 # use only few cycles for illustrational purposes
mirt.wrapper.coef(res2$mirt)
summary(res2$mirt)
print(res2$mirt)

#############################################################################
# EXAMPLE 4: mirt conversions for polytomous dataset data.big5
#############################################################################

data(data.big5)
# select some items
items <- c( grep( "O", colnames(data.big5), value=TRUE )[1:6],
     grep( "N", colnames(data.big5), value=TRUE )[1:4] )
# O3 O8 O13 O18 O23 O28 N1 N6 N11 N16
dat <- data.big5[, items ]
library(psych)
psych::describe(dat)

library(TAM)
#******************
#*** Model 1: Partial credit model in TAM
mod1 <- TAM::tam.mml( dat[,1:6] )
summary(mod1)
# convert to mirt object
mmod1 <- sirt::tam2mirt( mod1 )
rmod1 <- mmod1$mirt
# coefficients in mirt
coef(rmod1)
mirt.wrapper.coef(rmod1)
# model fit
mirt::M2(rmod1)
# item fit
mirt::itemfit(rmod1)
# plots
plot(rmod1,type="trace")
plot(rmod1, type="trace", which.items=1:4 )
mirt::itemplot(rmod1,"O3")

#******************
#*** Model 2: Generalized partial credit model in TAM
mod2 <- TAM::tam.mml.2pl( dat[,1:6], irtmodel="GPCM" )
summary(mod2)
# convert to mirt object
mmod2 <- sirt::tam2mirt( mod2 )
rmod2 <- mmod2$mirt
# coefficients in mirt
mirt.wrapper.coef(rmod2)
# model fit
mirt::M2(rmod2)
# item fit
mirt::itemfit(rmod2)

## End(Not run)

sirt documentation built on May 29, 2024, 8:43 a.m.