Interpreting coefficients with exploratory Extended Mixed-Effect Item Response Model

[Seongho Bae]

Hi, Phil

I want to know how can I Interpret coefficients with exploratory Extended Mixed-Effect Item Response Model.

1. We're all coefficients rotated with model > 1?
   
2. If not, How to interpret exploratory coefficients?

Seongho

[Phil Chalmers]

The exploratory model would be hard to interpret here because the orientation of the loadings is arbitrary (fixed at a particular orthogonal configuration during estimation). You could probably interpret the fixed-effect intercept terms as is, but loading coefficients are subject to post-rotation. In general, it's better to use this methodology with known factor structure configurations.

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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Seongho Bae]

Okay, thanks for making it clear.

If I don't know any factor structures of newly constructing test with clustered data structure (as known as 'dual local dependence'), Could you give any tips for investigating the factor structure?

Seongho

2017년 8월 4일 금요일 오후 10시 32분 46초 UTC+9, Phil Chalmers 님의 말:

Phil

To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package...@googlegroups.com.

[Phil Chalmers]

The standard exploratory factor analysis strategy is reasonable here, as well as testing more specific confirmatory models. Leaving out explanatory predictor variables doesn't typically influence the structure of the latent traits (unless the structure is something you're trying to explain with external covariates). 

Phil

To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package+unsubscribe@googlegroups.com.

[Seongho Bae]

Oh, that means ignoring random effects clusters does not influence to structure of the latent traits! That sounds cool!

Can I understand to adding explanatory variables just do controlling standard error of measurement, not factor structure like the SEM perspectives; even the explanatory variables were cross-classified random effect variables?

Seongho

2017년 8월 4일 금요일 오후 10시 55분 6초 UTC+9, Phil Chalmers 님의 말:

[Phil Chalmers]

The crossed-effect random effects could affect the model, because this allows the random grouping variables to have different random factor loadings depending on group membership. So, just be sure you're careful with the structure and population you wish to infer to a priori (there's no hard-fast rule here).  

Phil

On Fri, Aug 4, 2017 at 10:02 AM, Seongho Bae <seongh...@gmail.com> wrote:

Oh, that means ignoring random effects clusters do not influence to structure of the latent traits! That sounds cool; Can I understand adding explanatory variables controlling standard error of measurement, not factor structure even the explanatory variables were cross-classified random effect variables?

To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package+unsubscribe@googlegroups.com.

[Seongho Bae]

Okay, if there's no hard-fast rule here, How about you think for set the alternative method?

1. Extracting two models with or without random effects in the mixedmirt()
2. Differ model fits among both by AIC, BIC, DIC families
3. If random effects model was better, extract item parameters with modvalues() and to apply mirt() them; if not, recalibrate with mirt(), without any covariates.

Seongho

2017년 8월 5일 토요일 오전 3시 54분 46초 UTC+9, Phil Chalmers 님의 말:

[Phil Chalmers]

Sure, that sounds reasonable.

Phil

To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package+unsubscribe@googlegroups.com.

[Seongho Bae]

Okay, with your advice, I'm doing my experiment with build software to find the global-optimal model using exploratory and confirmatory factoring with considering cross-classified structure.

Always Thanks for your kindly pieces of advice.

Best,

Seongho

2017년 8월 5일 토요일 오후 11시 3분 18초 UTC+9, Phil Chalmers 님의 말: