mirt (version 0.5.0)

wald: Wald test for mirt models

Description

Compute a Wald test given an L vector or matrix of numeric contrasts. Requires that the model information matrix be computed (including SE = TRUE when using the EM method). Use wald(model) to observe how the information matrix columns are named, especially if the estimated model contains constrained parameters (e.g., 1PL). The information matrix names are labelled according to which parameter number(s) they correspeond to (to check the original numbering use the option pars = 'values' in the original estimation function).

Usage

wald(object, L, C = 0)
  ## S3 method for class 'wald':
print(x, ...)

Arguments

L

a coefficient matrix with dimensions nconstrasts x npars, or a vector if only one set of contrasts is being tested. Ommiting this value will return the column names of the information matrix used to identify the (potentially constrained) parameter

object

estimated object from mirt, bfactor, confmirt, multipleGroup, or mixedmirt

C

a constant vector/matrix to be compared along side L

x

an object of class 'wald'

...

additional arguments to be passed

Examples

#View parnumber index
data(LSAT7)
data <- expand.table(LSAT7)
cmodel <- confmirt.model()
   F1 = 1,4,5
   F2 = 2,3


mod <- mirt(data, cmodel, SE = TRUE)
coef(mod)

#see how the infomation matrix relates to estimated parameters, and how it lines up with the index
(infonames <- wald(mod))
index <- mirt(data, cmodel, pars = 'values')
index

#second factor slope equal to 0?
L <- rep(0, 10)
names(L) <- infonames
L[3] <- 1
wald(mod, L)

#simultaneously test equal factor slopes for item 2 and 3, and 4 and 5
L <- matrix(0, 2, 10)
colnames(L) <- infonames #colnames() not required
L[1,3] <- L[2, 7] <- 1
L[1,5] <- L[2, 9] <- -1
L
wald(mod, L)

#logLiklihood tests (requires estimating a new model)
mod2 <- mirt(data, cmodel, constrain = list(c(7,12), c(16,21)))
anova(mod2, mod)