1. the simulation of selection was based on the basis of X from the top percentile. This set let the range of selected ability only locate at high level. Why not to set the range of ability locate at the mediate level? In practice, the applicants often were interested in the reliability in the population of mediate level. Since the items often were design to have the difficulty matched to the concerned population, the concerned population located at mediate level in the test.
2. I don't understand why the w could inflate to over one. To avoid this problem, the author bounded the communality to be less than one. But see Function 2, w shouldn't be over one.
3. The reliability of range restriction based on IRT should be compared in this study. At least, IRT model was the true model in simulation, which is valuable to be investigated. This condition (IRT based reliability) also can be easily computed by using the reports of Conquest or WINSTEP software.