This study is said to employ a new approach for identifying item clusters. The proposed method is based on the hierarchical cluster analysis, together with the R statistics to assess the model fit. The proposed method does not assume any specified model underlies the analyzed item set, and uses a test statistic to determine the number of clusters.
Some questions about the study:
1) The basic structure of the procedure. The method is said to be similar to “partial hierarchical cluster analysis”, then what is the difference between “partial hierarchical cluster analysis” and “hierarchical cluster analysis”?
2) It seems the essential part is to define the f function, which used to assign items into different clusters, what is the f function?
4) For the simulation design, why the correlation between person would be in interest? What is the meaning of this design? How person correlation would affect the clustering of items?
4) Refer to the Table 1, what means by “no error”?
5) Comparing the results in Table 1 and Table 2, the proposed algorithm was better than the traditional method only under the condition of 1,000 persons, 10 items, and 0 correlation. It seems that the traditional method is more stable and powerful under most of the conditions. Then the proposed method could only be applied to some very strict conditions? What is the advantage of the proposed method?
6) The second paragraph in the Discussion part, the last sentence said the usefulness of the R statistic was assessed in the simulation. Is the R statistic be used in the traditional method? How to assess its usefulness in the simulation studies’ results?
7) The fifth paragraph in the Discussion part.
The proposed method performed worse if the sample size is small, and this result is similar to the study of the R statistic. Then whether it could be said as that, the performance of the R statistic affects the performance of the proposed algorithm?
What is the relationship between the f function in this method and the test statistic? How to choose an appropriate test statistic?