Can't the MHRM calculate information matrix in the model for MI estimation?

[Seongho Bae]

Dear Phil,

Can't the MHRM calculate information matrix in the model for MI estimation?

mod <- mirt(Science, 1, SE=TRUE)

fscores(mod, MI = 30) # work

mod <- mirt(Science, 1, SE=TRUE, method = 'QMCEM')

fscores(mod, MI = 30) # work

mod <- mirt(Science, 1, SE=TRUE, method = 'MHRM')

fscores(mod, MI = 30) # doesn't work

--

Seongho Bae

[Phil Chalmers]

The MHRM information matrix isn't always the most stable, try dropping the SEtol default like so to make it run longer and stabilize:

mod <- mirt(Science, 1, SE=TRUE, method = 'MHRM', technical = list(SEtol = 1e-4))

fscores(mod, MI=30)

Phil

--
You received this message because you are subscribed to the Google Groups "mirt-package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Seongho Bae]

Okay, Thanks for your advices. :)

Anyway, MHRM with dropping the SEtol 1e-5 were too slow. I was failed to get a stable result 1e-4 condition, I can not get any results even waiting 7 hours with just polytomous 5 items 1e-5 condition. Is this normal situation?

2015년 3월 27일 금요일 오전 8시 6분 14초 UTC+9, Phil Chalmers 님의 말:

[Phil Chalmers]

It's possible, but it's also a function of how stable the information matrix is in the first place. More unstable (i.e., a flat log-likelihood function) models take a lot longer to stabilize, so if your model is very complicated and the sample size is smaller, then you may have great difficulty in reaching a stable estimate.

Phil