mirt (version 1.17.1)

lagrange: Lagrange test for freeing parameters

Description

Lagrange (i.e., score) test to test whether parameters should be freed from a more constrained baseline model. The required derivative terms are evaluated numerically from the numerical_deriv function. Defining a mirtCluster will allow the independent tests to be run in parallel.

Usage

lagrange(mod, parnum, type = "forward", ...)

Arguments

mod

an estimated model

parnum

a vector, or list of vectors, containing one or more parameter locations/sets of locations to be tested. See objects returned from mod2values for the locations

type

type of numerical derivatives to use to find the associated gradient terms. Default is 'forward', but can also be 'Richardson' and 'central' (see numerical_deriv)

...

additional arguments to pass to numerical_deriv

See Also

wald

Examples

dat <- expand.table(LSAT7)
mod <- mirt(dat, 1, 'Rasch')
(values <- mod2values(mod))

#test all slopes individually
parnum <- values$parnum[values$name == 'a1']
lagrange(mod, parnum)

# compare to LR test for first two slopes
mod2 <- mirt(dat, 'F = 1-5
                   FREE = (1, a1)', 'Rasch')
coef(mod2, simplify=TRUE)$items
anova(mod, mod2)

mod2 <- mirt(dat, 'F = 1-5
                   FREE = (2, a1)', 'Rasch')
coef(mod2, simplify=TRUE)$items
anova(mod, mod2)

# test slopes first two slopes and last three slopes jointly
lagrange(mod, list(parnum[1:2], parnum[3:5]))

# DIF test
set.seed(1234)
n <- 30
N <- 500

a <- matrix(1, n)
d <- matrix(rnorm(n), n)
group <- c(rep('Group_1', N), rep('Group_2', N))

# groups completely equal
dat1 <- simdata(a, d, N, itemtype = 'dich')
dat2 <- simdata(a, d, N, itemtype = 'dich')
dat <- rbind(dat1, dat2)
mod <- multipleGroup(dat, 1, group=group,
                     invariance=c('free_means', 'free_var', colnames(dat)))
coef(mod, simplify=TRUE)
values <- mod2values(mod)

# mirtCluster()
lagrange(mod, list(c(123, 124), c(239,240)))