How to model multiple latent variables of a set of items using different coding schemes?

Question

I am using the mirt package of R and would like to test a multidimensional PCM or nominal model on 9 items. That is, I want to use the original rating scores (from 0 to 4) and an alternative coding of the same items (eg, 0/1).

I specify that the first latent variable F1 be modelled using the first 9 columns of the data matrix (the original ratings) and the second latent variable be modelled based on the next 9 columns:

model=mirt.model('F1=1-9
                  F2=10-18')

Logically, the package then computes a latent variable F1 for items 1 to 9 with zero loadings for F2 and latent variable for F2 based on items 10 to 18 with zero loadings on F1, as if columns 10 to 18 were separate items, which is not what I need.

Any ideas how to force mirt think that columns 10 to 18 are actually the same items as 1 to 9?

Many thanks! KH

Answer 1

Apparently, this is not possible either. However, it can be implemented in the TAM package. See Implement ConQuest score command in TAM for the respective code.