Item response model in Stata 12

[Maya Lani]

Dear all,

I would like to use item response model (that is developed with Stata version 14) in Stata 12.

However, whenever I tried to install it via "ssc install itr" or " findit itr" I could not find it.

Is it not possible to use item response model with Stata 12?

P.S: the command for the item response model is
irt grm

Thank you,
Maya

[Mike Lacy]

Stata 14 has a built-in command called -irt-. If this is what you want, I think you're out of luck for getting help, as I don't think people would want to help you get this "for free" even if this were possible. If you actually did mean -itr- , the fact that -findit- does not reveal it means that no such command is present in any of the sources that -findit- can see. Perhaps you know of some non-public module called -itr-, in which case you would need to contact the author. I presume this is not the case.

Some user-written IRT procedures are available for you to install, and they likely will run under Stata 12: Try -findit rasch- to list some of them and see if they are of interest.

[Rich Goldstein]

I assume what you typed above in quotes contains typos - but correcting those won't help anyway as IRT was added in Stata 14; however, if you type "hsearch irt" and scroll down you will find at least two user-written commands that, I think, will run in Stata 12: geekel2d and simirt; also version 12 has SEM and some IRT models can be estimated via SEM; I have not looked at the version 12 manuals to see if IRT is discussed there but you should do that

[Maya Lani]

Mike Lacy Dear Mike, I am aware that the ITR models are developed only for Stata 14 but I was hoping there would be a user written command for Stata 12 that I can use. Anyhow, thank you for your help.

[Maya Lani]

Rich Goldstein The manual for IRT model itself only mentions Stata 14. I will look into the user written commands you mentioned and see if I can use them in Stata 12. Thank you so much for your response! much appreciated.

[Oded Mcdossi]

Maya Lani You can estimate some of IRT models using gllamm. The learning curve of gllamm is too steep but if you don't have any chance to find Stata 14, then it worth it.

[Maya Lani]

Oded Mcdossi I am afraid I cannot get my hands on Stata 14. I will look into gllamm models. Thank you.

[Oded Mcdossi]

Here are some examples:
http://www.gllamm.org/examples.html
http://www.gllamm.org/aggression.html
http://www.gllamm.org/books/irt.html

[Maya Lani]

Oded Mcdossi thank you very much for taking time and providing me with examples.

it might sound like a silly question but i am new to these kind of models...is the gllamm applicable with rank data?
As an example of my data, each respondent was asked to rate (on a scale 0 to 10) the importance of 5 different subjects (maths, biology, geography, psychology and economics) to his education.

[Oded Mcdossi]

Your question is not silly at all. I think gllamm can estimate this model, but you should first choose the right model under the IRT umbrella and then check if the program can estimate that. I can't say much without more details about your aims and I'm not completely sure about that, but with some coding of your scale (for example recode the answers to binary like scale>6=1 otherwise 0), if this applicable, I think you can estimate your model within the Rasch approach (read this tutorial for more details: http://www.stata.com/support/faqs/st...s/rasch-model/).

[Maya Lani]

Thank you very much for your advice. I am now looking into the appropriate IRT and how it is applicable with gllam.

I already looked into the Rasch model but I think I cannot really use it because the Rasch model involves a person's ability and the difficulty of the question.
I can estimate a person's ability but the difficulty factor (e.g. the difficulty of the questions asked) I don't see how I can estimate...