Robust Clustering on Rule Usage of Probability Reasoning with Raw Rule Score

The purpose of this study is to analyze the rule usage of probability reasoning items by a robust clustering method proposed by Yang and Wu (2004). Based on raw rule scoring of four rules, each subject has four scores as to rule usage. The clustering results show that the optimal group number is four and each group corresponds to a unique rule usage. Thanks to the robustness, patterns of rule usage for each cluster display the cognitive structures of those who belong to that group. It also shows that patterns of rule usage vary with different grade of subjects, but not with gender.