search

philchalmers / mirt

Multidimensional item response theory
https://philchalmers.github.io/mirt/
201 stars 75 forks source link

complete missing and fscore for multigroup models #252

Closed Keddeko closed 4 months ago

Keddeko commented 4 months ago

Hi there I have a dataset with design based complete missings for some rows. Instead of merging with a PID or similar, I have no problem getting the ability estimates with fscores(mirt_model, full.scores.SE = TRUE) for the full matrix with complete NAs for some rows, but when I make a multigroup model with multipleGroup and try to retrieve the results with full.scores.SE or full.scores in the same way I get the following: [<-(*tmp*, completely_missing, , value = NA) : subscript out of bounds

If I on the other hand make full row deletion for the observations with complete missings it works fine. But, should the fscores(full.scores=TRUE) not work for both the multipleGroup and for mirt objects also in cases with complete missings?

It would be nice in order to have a full dataset (with NAs naturally for rows with complete missings). Best and thanks for a cool and well documented package ! ' Christian

philchalmers commented 4 months ago

Hi Christian,

Thanks for the report. Does this issue still exist on the current dev version here on Github? If so, could you please provide a brief reprex demonstrating the issue so that it can be patched? Thanks in advance.

Keddeko commented 4 months ago

Dear Phil I have tried the current dev verion and it solved the issue - thanks Best regards Christian