The purpose of this study is that to evaluate the performances of different scaling method when sample size is small or test length is short. 5 scaling methods * 2 IRT models * 5 sample sizes * 4 test lengths resulted in a total of 200 conditions.
Comments & Questions:
1. A lot of conditions were examined in this study but only a part of them was presented. That is the tables or figures only show summary results for all conditions, some informations maybe ignored.
2. I don’t know what situation is agreed to the simulation that is test is short test length and small sample size in practice.
3. In overall, the concurrent scaling method seems outperform to others in small sample size and short test length. But the authors say that the parameters estimation of separate scaling methods maybe influenced by the short test length which has small number of common items. If we would like to examine whether the concurrent scaling method is recommend to small sample size and short test length. I don’t sure if we can use it in a condition of small sample size but longer test length included to eliminate the effect of confounding factor.