longitudinal mirt

[Makrem Ben Youssef]

Hi Phil

 

I've constructed a measurement index composed of 30 binary variables for which I have an unbalanced panel of 281- firm-year observations. I want to check for index validity using mirt model but i'm wondering if ther's a special treatment for longtidunal data in mirt especially slow-changing variables. Alternatively, is it wrong to suppose that the observations are cross-sectional and run ordinary mirt model?

 

Tanhk you for your help!!

Makrem

[Phil Chalmers]

Hi Makrem,

I'm not entirely sure I understand without getting more information, but yes you can do longitudinal IRT with mirt (an older example is on the wiki, which should probably be updated given the newer syntax arguments). There's nothing wrong with looking at cross sections within each time block, but over time chunks I wouldn't recommend it. 

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.

[Makrem Ben Youssef]

Hi Phil

Thank you for your quick answer, I saw the example on the wiki, but it seems that only one dimension is required for each time period which is not my case, i need to define two factors for each year, is it possible?

Thank you for your precious support!

Makrem

[Phil Chalmers]

Of course, you are free to add more factors to the design within each time point, however you probably will hit a limit in the number of factors to include eventually because the number of dimensions may grow to be too large. But adding one more factor in and constraining the slopes to be equal across time points shouldn't be too bad. Cheers.

Phil 

--

[Makrem Ben Youssef]

Hi Phil

I go back to the example you gave and try to replace bfactor by mirt (because it's the most appropriate model for my case) but it dosen't work, Besides there are too many steps in the code that i don't really undrestand such as: setup within time constraints and create constrain list, Are these steps complusory in longitudinal mirt??

THe code I used i something like this 

datapanel<- mirt.model ('Time1= 1-15,16-27

Time2= 28-42,43-54

Time3= 55-69,70-82

COV= Time1*Time2*Time3,Time2*Time2,Time3*Time3')

modpanel<- mirt(datapanel,mirtmodel)

I know it seems  totally wrong but it's my first trial with longitudinal irt and tried to make analogy with simple mirt model

Please, any help would really appreciated

Thank you Phil

Markrem

Le jeudi 14 mai 2015 15:02:57 UTC+2, Makrem Ben Youssef a écrit :

[Phil Chalmers]

On Fri, May 15, 2015 at 8:24 AM, Makrem Ben Youssef <makrem.b...@googlemail.com> wrote:

Hi Phil

I go back to the example you gave and try to replace bfactor by mirt (because it's the most appropriate model for my case) but it dosen't work, Besides there are too many steps in the code that i don't really undrestand such as: setup within time constraints and create constrain list, Are these steps complusory in longitudinal mirt??

THe code I used i something like this 

datapanel<- mirt.model ('Time1= 1-15,16-27

Time2= 28-42,43-54

Time3= 55-69,70-82

COV= Time1*Time2*Time3,Time2*Time2,Time3*Time3')

modpanel<- mirt(datapanel,mirtmodel)

I know it seems  totally wrong but it's my first trial with longitudinal irt and tried to make analogy with simple mirt model

This seems fine, just be sure to set up the item-level equality constraints as well across each time point to identify the model. I generally prefer the random slope intercept form shown in the bfactor() example, but this way is fine too.

Phil

 

Please, any help would really appreciated

Thank you Phil

Markrem

Le jeudi 14 mai 2015 15:02:57 UTC+2, Makrem Ben Youssef a écrit :

Hi Phil

 

I've constructed a measurement index composed of 30 binary variables for which I have an unbalanced panel of 281- firm-year observations. I want to check for index validity using mirt model but i'm wondering if ther's a special treatment for longtidunal data in mirt especially slow-changing variables. Alternatively, is it wrong to suppose that the observations are cross-sectional and run ordinary mirt model?

 

Tanhk you for your help!!

Makrem

--

[Makrem Ben Youssef]

Hi Phil

excuse my ignorance but is this the right syntax to constrain all item slopes to be equal

modelpanel<- mirt.model ('Time1= 1-15,16-27

 +CONSTRAIN (1-15,16-27,a1)

+ Time2= 28-42,43-54

+CONSTRAIN (28-42,43-54,a1)

+ Time3= 55-69,70-81

+CONSTRAIN (55-69,70-81,a1)

+ COV= Time1*Time2*Time3,Time2*Time2,Time3*Time3')

I'm really confused, I don't know how to identify the model

Thank you for your patience

Le jeudi 14 mai 2015 15:02:57 UTC+2, Makrem Ben Youssef a écrit :