generate.mirt_object function

Generate a mirt object from population parameters

This function generates a mirt object from known population parameters, which is then passed to mirtCAT for running CAT applications.

generate.mirt_object(
  parameters,
  itemtype,
  latent_means = NULL,
  latent_covariance = NULL,
  key = NULL,
  min_category = rep(0L, length(itemtype))
)

Arguments

• parameters: a matrix or data.frame of parameters corresponding to the model definitions listed in mirt. Each row represents a unique item, while the column names correspond to the respective parameter names. If a parameter is not relevant for a particular item/row then use NA's as placeholders
• itemtype: a character vector indicating the type of item with which the parameters refer. See the itemtype argument in mirt. Note that this input is only used to determine the relevant item class for the rows in parameters, therefore many inputs are interchangeable (e.g., '2PL' generates the same internal model object as '3PL'). If only a single value is provided then all items types will be assumed identical
• latent_means: (optional) a numeric vector used to define the population latent mean structure. By default the mean structure is centered at a 0 vector
• latent_covariance: (optional) a matrix used to define the population variance-covariance structure between the latent traits. By default the relationship is assumed to be orthogonal standard normal (i.e., an identity matrix)
• key: scoring key required for nested-logit models. See mirt for details
• min_category: the value representing the lowest category index. By default this is 0, therefore the response suitable for the first category is 0, second is 1, and so on up to K - 1

Examples

## Not run:

### build a unidimensional test with all 3PL items

nitems <- 50
a1 <- rlnorm(nitems, .2,.2)
d <- rnorm(nitems)
g <- rbeta(nitems, 20, 80)

pars <- data.frame(a1=a1, d=d, g=g)
head(pars)

obj <- generate.mirt_object(pars, '3PL')
coef(obj, simplify = TRUE)
plot(obj, type = 'trace')

### build a two-dimensional test  
## all graded items with 5 response categories

nitems <- 30
as <- matrix(rlnorm(nitems*2, .2, .2), nitems)
diffs <- t(apply(matrix(runif(nitems*4, .3, 1), nitems), 1, cumsum)) 
diffs <- -(diffs - rowMeans(diffs)) 
ds <- diffs + rnorm(nitems)
pars2 <- data.frame(as, ds)
colnames(pars2) <- c('a1', 'a2', paste0('d', 1:4))
head(pars2)

obj <- generate.mirt_object(pars2, 'graded')
coef(obj, simplify = TRUE)

### unidimensional mixed-item test

library(plyr)
pars3 <- rbind.fill(pars, pars2) #notice the NA's where parameters do not exist
obj <- generate.mirt_object(pars3, itemtype = c(rep('2PL', 50), rep('graded', 30)))
coef(obj)
itemplot(obj, 51)
itemplot(obj, 1, drop.zeros=TRUE)
## End(Not run)

References

Chalmers, R., P. (2012). mirt: A Multidimensional Item Response Theory Package for the R Environment. Journal of Statistical Software, 48(6), 1-29. tools:::Rd_expr_doi("10.18637/jss.v048.i06")

Chalmers, R. P. (2016). Generating Adaptive and Non-Adaptive Test Interfaces for Multidimensional Item Response Theory Applications. Journal of Statistical Software, 71(5), 1-39. tools:::Rd_expr_doi("10.18637/jss.v071.i05")

See Also

mirt, mirtCAT, generate_pattern

Author(s)

Phil Chalmers rphilip.chalmers@gmail.com