Scoring (bypassing "only one response category" error)

[Marcus Waldman]

Hi!

I've calibrated item parameters and build an large item bank using `mirt.` Now I would like to use the `mirt::fscores` function to score on new item response data that is separate from the calibration sample.

Here's the general way I was hoping to go about it:

1. Extract the parameter estimates from the fitted model :

`fit_calibration = mirt::mirt(..., data = calibration_data)

pars_calibration = mirt::mod2values(fit_calibration)`

2. Set `pars_calibration$est = F` for all parameters and pass to a new mirt model object with the new data

`fit_new = mirt::mirt(..., data = new_data, pars = pars_calibration) `

3. Then use mirt::fscores(fit_new) to get EAP scores

Unfortunately, I get stuck on #2 --trying to make a new mirt object that "sees" the new data. This is because not all items are answered, for example and some items will have no variance/not all response categories will be represented. Here is an example error:

` Error: The following items have only one response category and cannot be estimated: CC17 CC5 CC6 CC9`

Obviously it is true that this would be problematic if I were trying to freely estimate item parameters with the new data. But all item parameters are constrained to their calibrated value, so I'm wondering if there is a way to bypass this error message and create the necessary model to pass to `mirt::fscores`.

Thanks for your help!