MIRT with small sample size but valid models

Hi there,

I already posted this question to the google mirt group but didn't get an answer, maybe there are some IRT experts here who can help me.

I did some IRT analysis (mostly unidimensional GRM) on data with a sample size of around N=100. After some model and item wrangling I got valid models (RMSEA around 0.5, CFI and TLI > 0.93, sometimes even 0.97, even Chi2 was ok). Data reliability (empirical) rxx is > 0.7.

All articles I found about minimal sample size for IRT suggest at least N=500. Can my data be valid nontheless or do I have to use other methods?

I wanted to use GRM in the first place to analyse the validity of the item levels, which indeed gave me a lot of information about my code system, but now I'm concerned about the calculated factor scores.

Thanks a lot for your help!


This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.