Comparing results of eRm and mirt

[Jeanette Müller]

Dear mirt-developer,

first of all - thank you very much for your package!
I recently tried to compare the results of eRm and mirt for a partial credit model.
So I used the "Rasch" command and tried my best - but I couldn´t map the results - could you help me out?
It would be very helpful for me if one of you experts could write a little "demo" syntax or tell me where I took the "wrong" turn...

Thank you very much in advance,
best regards

Jeanette

[Phil Chalmers]

Hi Jeanette,

The packages use different parameterizations for the models, so direct
comparison between them would require a little math to determine the
link for partial credit models. For instance, we can compare the Rasch
models between the packages like so:

library(eRm)
library(mirt)
dat <- expand.table(LSAT7)
res <- RM(dat)
cfs <- coef(res)
scale(cfs)

#for 'Rasch', fix the slopes = 1, estimate all the b's and latent
variance parameter
mres <- mirt(dat, 1, itemtype= 'Rasch')
cfs <- coef(mres, IRTpars = TRUE)
b <- numeric(5)
for(i in 1:5) b[i] <- cfs[[i]][2L]
scale(b)

You'll notice that you have to rescale the results of the intercepts
to put them onto the same metric. They also won't be exactly the same
since eRm uses CML estimation while mirt uses MML.

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/groups/opt_out.

[Jeanette Müller]

Hi Phil,

thank you very much for the quick and helpful answer!
If I understand you correct it would be possible but a bit more of code to do the same "trick" for the PCM?

I have another question but most probably it is better to open a new thread.

Best Jeanette

[Phil Chalmers]

Hi Jeanette,

On Thu, Jan 9, 2014 at 12:37 PM, Jeanette Müller <petepa...@gmail.com> wrote:
> Hi Phil,
>
> thank you very much for the quick and helpful answer!
> If I understand you correct it would be possible but a bit more of code to
> do the same "trick" for the PCM?

Yes it should be possible, I just don't know the exact
transformational relationship between the parameterizations off hand.
It would require a bit of worked math to determine this, but I have no
doubts that a direct relationship exists.

Phil

[Jeanette Müller]

Hi Phil,

once again thank you very much for the quick and helpful answer!

Best

Jeanette