polytomous Rasch model in snowIRT module

[RMD]

Hello,
I'm new to running a polytomous Rasch model with snowIRT and have a couple of questions:

- Is the MADaQ3 always an appropriate as a measure of model fit? Are there other options? (And do I really need to test model fit if I'm focussing on item infit and outfit)? The item fit stats seem very good but the MADaQ3 is significant.

- What does it mean if under 'item statistics of the rating scale model' all of the values under 'Measure' are negative? (I'm assuming this 'Measure' is the same as the 'item difficulty' - right?)

Many thanks for any pointers!

[seol]

Hi RMD

you can use snowRMM::Rasch model by downloading it in jamovi library and Measure is logit scale(log odds unit).

Cheers
Seol

[RMD]

Many thanks Seol, I'll check out snowRMM.

[RMD]

I actually tried a polytomous Rasch model with snowRMM and compared the results to a polytomous Rasch model with snowIRT and the item fit statistics look quite different.

Here are the two differing results:
https://1drv.ms/b/s!Ak18NDjxZnZr1hiPW0i ... -?e=R3JFcp

What would explain such a difference? I don't know which would be better to report (It's a 5 point likert scale on the frequency of a an experience).

Also, is there a way within these packages of checking for unidimensionality? I understand that doing a PCA on the residuals is not included, right?

Many thanks for any ideas!

[seol]

HI RMD
Sorry to delay response. I have been using TAM R package inside snowIRT module. On the other hand, snowRMM using mixRasch R package. That is, TAM is using Marginal maximum likelihood estimates to get item logit score. and mixRasch is using Joint Maximum likelihood estimation.

Cheers
Seol

[RMD]

Oh, I see. That makes sense. Thanks Seol!
Russell