sirt (version 1.5-0)

data.wide2long: Converting a Data Frame from Wide Format in a Long Format

Description

Converts a data frame in wide format into long format.

Usage

data.wide2long(dat, id = NULL, X = NULL, Q = NULL)

Arguments

dat

Data frame with item responses and a person identifier if id != NULL.

id

An optional string with the variable name of the person identifier.

X

Data frame with person covariates for inclusion in the data frame of long format

Q

Data frame with item predictors. Item labels must be included as a column named by "item".

Value

• Data frame in long format

Examples

#############################################################################
# EXAMPLE 1: data.pisaRead
#############################################################################
library(lme4)

data(data.pisaRead) 
dat <- data.pisaRead$data 
Q <- data.pisaRead$item   # item predictors

# define items
items <- colnames(dat)[ substring( colnames(dat) , 1 , 1 ) == "R" ]
dat1 <- dat[ , c( "idstud" , items ) ]
# matrix with person predictors
X <- dat[ , c("idschool" , "hisei" , "female" , "migra") ]

# create dataset in long format
dat.long <- data.wide2long( dat=dat1 , id="idstud" , X=X , Q=Q )

#***
# Model 1: Rasch model
mod1 <- lme4::glmer( resp ~ 0 + ( 1 | idstud ) + as.factor(item) , data = dat.long ,
            family="binomial" , verbose=TRUE)
summary(mod1)

#***
# Model 2: Rasch model and inclusion of person predictors
mod2 <- lme4::glmer( resp ~ 0 + ( 1 | idstud ) + as.factor(item) + female + hisei + migra,
           data = dat.long , family="binomial" , verbose=TRUE)
summary(mod2)

#***
# Model 3: LLTM
mod3 <- lme4::glmer(resp ~ (1|idstud) + as.factor(ItemFormat) + as.factor(TextType), 
            data = dat.long , family="binomial" , verbose=TRUE)
summary(mod3)

#############################################################################
# SIMULATED EXAMPLE 2: Rasch model in lme4
#############################################################################

set.seed(765)
N <- 1000  # number of persons
I <- 10    # number of items
b <- seq(-2,2,length=I)
dat <- sirt::sim.raschtype( rnorm(N,sd=1.2) , b=b )
dat.long <- data.wide2long( dat=dat )
#***
# estimate Rasch model with lmer
library(lme4)
mod1 <- lme4::glmer( resp ~ 0 + as.factor( item ) + ( 1 | id_index) , data = dat.long ,
             verbose=TRUE , family="binomial")
summary(mod1)
  ##   Random effects:
  ##    Groups   Name        Variance Std.Dev.
  ##    id_index (Intercept) 1.454    1.206   
  ##   Number of obs: 10000, groups: id_index, 1000
  ##   
  ##   Fixed effects:
  ##                        Estimate Std. Error z value Pr(>|z|)    
  ##   as.factor(item)I0001  2.16365    0.10541  20.527  < 2e-16 ***
  ##   as.factor(item)I0002  1.66437    0.09400  17.706  < 2e-16 ***
  ##   as.factor(item)I0003  1.21816    0.08700  14.002  < 2e-16 ***
  ##   as.factor(item)I0004  0.68611    0.08184   8.383  < 2e-16 ***
  ##   [...]