There are essentially three models: ‘unidim’, ‘covariance’, and ‘factor’.
‘unidim’ analyzes a single item. ‘covariance’ is suitable for two or more items.
Once you have vetted your items with the ‘unidim’ and ‘covariance’ models,
then you can try the ‘factor’ model.
For each model, there is a ‘_ll’ variation. This model
includes row-wise log likelihoods suitable for feeding to loo
for efficient approximate leave-one-out cross-validation.
There is also a special model ‘unidim_adapt’. Except for
this model, the other models require a scaling constant. To find
an appropriate scaling constant, we recommend fitting
‘unidim_adapt’ to each item separately and then take the
median of median point estimates to set the scale. ‘unidim_adapt’ requires a
varCorrection constant. In general, a varCorrection of 2.0 or 3.0
should provide optimal results.