Convert a data.frame or tibble with factor variables (items) to integers,
keeping the original factor levels (i.e. response categories) and correct
answers (stored as an key attribute of each item) alongside.
nominal_to_int(Data, key)List of original levels with logical attribute key, which stores
the information on which response (level) is considered correct. Note
that levels not used in the original data are dropped.
data.frame or tibble with all columns being factors. Support for matrix is limited and behavior not guaranteed.
A single-column data.frame, (not matrix) tibble or -
preferably - a factor vector of levels considered as correct responses.
Fitting a nominal model using mirt::mirt() package requires the dataset to
consist only of integers, arbitrarily representing the response categories.
You can convert your dataset to integers on your own in that case.
On the other hand, BLIS model (and thus also the BLIRT parametrization)
further requires the information of correct item response category. On top of
that, the same information is leveraged when fitting a mirt model that
conserves the "directionality" of estimated latent ability (using a model
definition from obtain_nrm_def()). In these cases, you are recommended to
use nominal_to_int() (note that fit_blis() and blis() does this
internally). Note also that fitted BLIS model (of class BlisClass) stores
the original levels with correct answer key in its orig_levels slot,
accessible by a user via get_orig_levels().
Other BLIS/BLIRT related:
BlisClass-class,
coef,BlisClass-method,
fit_blis(),
get_orig_levels(),
obtain_nrm_def(),
print.blis_coefs()