R2conquest: Running ConQuest From Within R

Description

The function R2conquest runs the IRT software ConQuest (Wu, Adams, Wilson & Haldane, 2007) from within R.

Other functions are utility functions for reading item parameters, plausible values or person-item maps.

Usage

R2conquest(dat, path.conquest, conquest.name="console", converge=0.001,
    deviancechange=1e-04, iter=800, nodes=20, minnode=-6, maxnode=6,
    show.conquestoutput=FALSE, name="rasch", pid=1:(nrow(dat)), wgt=NULL, X=NULL,
    set.constraints=NULL, model="item", regression=NULL,
    itemcodes=seq(0,max(dat,na.rm=TRUE)), constraints=NULL, digits=5, onlysyntax=FALSE,
    qmatrix=NULL, import.regression=NULL, anchor.regression=NULL,
    anchor.covariance=NULL, pv=TRUE, designmatrix=NULL, only.calibration=FALSE,
    init_parameters=NULL, n_plausible=10,  persons.elim=TRUE, est.wle=TRUE,
    save.bat=TRUE, use.bat=FALSE, read.output=TRUE, ignore.pid=FALSE)
# S3 method for R2conquest
summary(object, ...)
# read all terms in a show file or only some terms
read.show(showfile)
read.show.term(showfile, term)
# read regression parameters in a show file
read.show.regression(showfile)
# read unidimensional plausible values form a pv file
read.pv(pvfile, npv=5)
# read multidimensional plausible values
read.multidimpv(pvfile, ndim, npv=5)
# read person-item map
read.pimap(showfile)

Value

A list with several entries

item: Data frame with item parameters and item statistics
person: Data frame with person parameters
shw.itemparameter: ConQuest output table for item parameters
shw.regrparameter: ConQuest output table for regression parameters
...: More values

Arguments

dat: Data frame of item responses
path.conquest: Directory where the ConQuest executable file is located
conquest.name: Name of the ConQuest executable.
converge: Maximal change in parameters
deviancechange: Maximal change in deviance
iter: Maximum number of iterations
nodes: Number of nodes for integration
minnode: Minimum value of discrete grid of \(\theta\) nodes
maxnode: Maximum value of discrete grid of \(\theta\) nodes
show.conquestoutput: Show ConQuest run log file on console?
name: Name of the output files. The default is 'rasch'.
pid: Person identifier
wgt: Vector of person weights
X: Matrix of covariates for the latent regression model (e.g. gender, socioeconomic status, ..) or for the item design (e.g. raters, booklets, ...)
set.constraints: This is the set.constraints in ConQuest. It can be "cases" (constraint for persons), "items" or "none"
model: Definition model statement. It can be for example "item+item*step" or "item+booklet+rater"
regression: The ConQuest regression statement (for example "gender+status")
itemcodes: Vector of valid codes for item responses. E.g. for partial credit data with at most 3 points it must be c(0,1,2,3).
constraints: Matrix of item parameter constraints. 1st column: Item names, 2nd column: Item parameters. It only works correctly for dichotomous data.
digits: Number of digits for covariates in the latent regression model
onlysyntax: Should only be ConQuest syntax generated?
qmatrix: Matrix of item loadings on dimensions in a multidimensional IRT model
import.regression: Name of an file with initial covariance parameters (follow the ConQuest specification rules!)
anchor.regression: Name of an file with anchored regression parameters
anchor.covariance: Name of an file with anchored covariance parameters (follow the ConQuest specification rules!)
pv: Draw plausible values?
designmatrix: Design matrix for item parameters (see the ConQuest manual)
only.calibration: Estimate only item parameters and not person parameters (no WLEs or plausible values are estimated)?
init_parameters: Name of an file with initial item parameters (follow the ConQuest specification rules!)
n_plausible: Number of plausible values
persons.elim: Eliminate persons with only missing item responses?
est.wle: Estimate weighted likelihood estimate?
save.bat: Save bat file?
use.bat: Run ConQuest from within R due a direct call via the system command (use.bat=FALSE) or via a system call of a bat file in the working directory (use.bat=TRUE)
read.output: Should ConQuest output files be processed? Default is TRUE.
ignore.pid: Logical indicating whether person identifiers (pid) should be processed in ConQuest input syntax.
object: Object of class R2conquest
showfile: A ConQuest show file (shw file)
term: Name of the term to be extracted in the show file
pvfile: File with plausible values
ndim: Number of dimensions
npv: Number of plausible values
...: Further arguments to be passed

Details

Consult the ConQuest manual (Wu et al., 2007) for specification details.

References

Wu, M. L., Adams, R. J., Wilson, M. R. & Haldane, S. (2007). ACER ConQuest Version 2.0. Mulgrave. https://shop.acer.edu.au/acer-shop/group/CON3.

Examples

Run this code

if (FALSE) {
# define ConQuest path
path.conquest <- "C:/Conquest/"

#############################################################################
# EXAMPLE 1: Dichotomous data (data.pisaMath)
#############################################################################
library(sirt)
data(data.pisaMath)
dat <- data.pisaMath$data

# select items
items <- colnames(dat)[ which( substring( colnames(dat), 1, 1)=="M" ) ]

#***
# Model 11: Rasch model
mod11 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
             pid=dat$idstud, name="mod11")
summary(mod11)
# read show file
shw11 <- sirt::read.show( "mod11.shw" )
# read person-item map
pi11 <- sirt::read.pimap(showfile="mod11.shw")

#***
# Model 12: Rasch model with fixed item difficulties (from Model 1)
mod12 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
             pid=dat$idstud, constraints=mod11$item[, c("item","itemdiff")],
             name="mod12")
summary(mod12)

#***
# Model 13: Latent regression model with predictors female, hisei and migra
mod13a <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
             pid=dat$idstud, X=dat[, c("female", "hisei", "migra") ],
             name="mod13a")
summary(mod13a)

# latent regression with a subset of predictors
mod13b <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
             pid=dat$idstud, X=dat[, c("female", "hisei", "migra") ],
             regression="hisei migra", name="mod13b")

#***
# Model 14: Differential item functioning (female)
mod14 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
             pid=dat$idstud, X=dat[, c("female"), drop=FALSE],
             model="item+female+item*female",  regression="",  name="mod14")

#############################################################################
# EXAMPLE 2: Polytomous data (data.Students)
#############################################################################
library(CDM)
data(data.Students)
dat <- data.Students

# select items
items <- grep.vec( "act", colnames(dat) )$x

#***
# Model 21: Partial credit model
mod21 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
              model="item+item*step",  name="mod21")

#***
# Model 22: Rating scale model
mod22 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
              model="item+step", name="mod22")

#***
# Model 23: Multidimensional model
items <- grep.vec( c("act", "sc" ), colnames(dat),  "OR" )$x
qmatrix <- matrix( 0, nrow=length(items), 2 )
qmatrix[1:5,1] <- 1
qmatrix[6:9,2] <- 1
mod23 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
            model="item+item*step", qmatrix=qmatrix, name="mod23")

#############################################################################
# EXAMPLE 3: Multi facet models (data.ratings1)
#############################################################################
library(sirt)
data(data.ratings1)
dat <- data.ratings1

items <- paste0("k",1:5)

# use numeric rater ID's
raters <- as.numeric( substring( paste( dat$rater ), 3 ) )

#***
# Model 31: Rater model 'item+item*step+rater'
mod31 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
              itemcodes=0:3, model="item+item*step+rater",
              pid=dat$idstud, X=data.frame("rater"=raters),
              regression="", name="mod31")

#***
# Model 32: Rater model 'item+item*step+rater+item*rater'
mod32 <- sirt::R2conquest(dat=dat[,items], path.conquest=path.conquest,
              model="item+item*step+rater+item*rater",
              pid=dat$idstud, X=data.frame("rater"=raters),
              regression="", name="mod32")
}

Run the code above in your browser using DataLab