Hi Phil and all,
I'm conducting an item analysis using Rasch model to examine item infit and outfit. My dataset contains missing values by design. Participants only administered a subset of items from a larger item pool, so it contains a lot of missing value. I compared three packages: mirt, TAM, and eRm, and using Winsteps as well. With TAM, eRm, and Winstep, I got the result, but with mirt I got error warning "Error: Only X2, G2, PV_Q1, PV_Q1*, X2*, and X2*_df can be computed with missing data." I'm not deep in the calculation behind it and I don't know exactly how this works for other packages but not for mirt. While comparing item difficulty (b), I got similar results from mirt and eRm. I knew those package used different estimations. Is that the reason why I couldn't get item fit with missing values in mirt?
One more question, if I want to create an adaptive test in mirtCAT, but I calibrate my test using other packages with a different estimation as mirt (e.g., eRm), will it cause a problem?
Thanks for your help
Hanif