1. It is confusing as sometime it said some statistics are unrealistic to obtain in real application, then sometime it said we can use some other statistics to replace. It seems calculating the CA estimation is not an easy job. And what is the criterion to evaluate whether the CA estimation is correct?
2. In this paper, all simulations assumed the true item parameters and model. Then in the discussion part, it said the CA estimation would be affected by model misfit. However, in reality, it is not possible to make such assumptions and get the true item parameters, in this case, how can an accurate CA be estimated?
3. When applying other methods for person classification, other than sum scores, or estimated ability, then whether the methods introduced still applicable?
4. Why should estimate CA? Use to evaluate the results of person classification?
If the purpose is to evaluate the results of person classification, why not use the SE of the estimated ability? Then the researcher would have a direct understanding. If a person’s ability is estimated with very small SE, this person can surely be correctly classified, no matter what classification criterion is used.