mirt (version 1.35.1)

PLCI.mirt: Compute profiled-likelihood (or posterior) confidence intervals

Description

Computes profiled-likelihood based confidence intervals. Supports the inclusion of equality constraints. Object returns the confidence intervals and whether the respective interval could be found.

Usage

PLCI.mirt(
  mod,
  parnum = NULL,
  alpha = 0.05,
  search_bound = TRUE,
  step = 0.5,
  lower = TRUE,
  upper = TRUE,
  inf2val = 30,
  NealeMiller = FALSE,
  ...
)

Arguments

mod

a converged mirt model

parnum

a numeric vector indicating which parameters to estimate. Use mod2values to determine parameter numbers. If NULL, all possible parameters are used

alpha

two-tailed alpha critical level

search_bound

logical; use a fixed grid of values around the ML estimate to determine more suitable optimization bounds? Using this has much better behaviour than setting fixed upper/lower bound values and searching from more extreme ends

step

magnitude of steps used when search_bound is TRUE. Smaller values create more points to search a suitable bound for (up to the lower bound value visible with mod2values). When upper/lower bounds are detected this value will be adjusted accordingly

lower

logical; search for the lower CI?

upper

logical; search for the upper CI?

inf2val

a numeric used to change parameter bounds which are infinity to a finite number. Decreasing this too much may not allow a suitable bound to be located. Default is 30

NealeMiller

logical; use the Neale and Miller 1997 approximation? Default is FALSE

...

additional arguments to pass to the estimation functions

References

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

Chalmers, R. P., Pek, J., & Liu, Y. (2017). Profile-likelihood Confidence Intervals in Item Response Theory Models. Multivariate Behavioral Research, 52, 533-550. 10.1080/00273171.2017.1329082

Neale, M. C. & Miller, M. B. (1997). The use of likelihood-based confidence intervals in genetic models. Behavior Genetics, 27, 113-120.

See Also

boot.mirt

Examples

# NOT RUN {
# }
# NOT RUN {
mirtCluster() #use all available cores to estimate CI's in parallel
dat <- expand.table(LSAT7)
mod <- mirt(dat, 1)

result <- PLCI.mirt(mod)
result

# model with constraints
mod <- mirt(dat, 'F = 1-5
                  CONSTRAIN = (1-5, a1)')

result <- PLCI.mirt(mod)
result

mod2 <- mirt(Science, 1)
result2 <- PLCI.mirt(mod2)
result2

#only estimate CI's slopes
sv <- mod2values(mod2)
parnum <- sv$parnum[sv$name == 'a1']
result3 <- PLCI.mirt(mod2, parnum)
result3

# }