1. 3PLM was used. it is well-known that c parameter is hard to estimate. I always doubt the results when sample size is small.
2. it is interesting to know what happens if large proportion of dif items exist.
3. table1 does not report the standard deviation over replications.
4. why larger sample size and items will lead to larger type 1 error when no adjusted corrections were used?
5. no accounts were given to explain why mh and Holmes methods benefited from corrections but not for irt based methods?
6. type 1 error seems underestimated when corrections methods were used. no accounts were given here.