This study investigate the rounding rules when conducting Angoff method in standard-setting. By comparing three kinds of rounding rules in the IRT model-based simulation, the authors stated that the more decimals keeping, the more accurate the judgement occur. The results seems still in a common sense which the more information inputs the more accurate the outputs if the true model was employed. However, I guess the most contribution of this study is trying to modeling the process of conducting Angoff method, just like the paper by Reckase (2006) before, although the authors still focus on the ideal scenario this time.
As the authors state that it is hard to know the intended cut-score in practice, it is also likely with the same logic to know about the cognitive properties of the panelists' sensitiveness about the rounding of item difficulty. The later issue is more important but it might depend on the contribution of cognitive psychologist rather than psychometricians. The concern of this issue also leading me to doubt about the implication to practitioners of the findings of this study.
On the other hand, as Prof. Wang suggested before, under the framework of this study, we can investigate other factors related to the process of standard-setting by incorporating the facet model. The possible factors can be but not limited to the effect of training and feedback to panelists, the interaction between panelists.