Scholarlink

This paper proposes a method for statistically assessing the similarity between two weight vectors, based on the Weights Similarity Coefficient (WSC). The WSC value measures the degree of similarity between two normalized vectors in the weight space. To determine the significance of this similarity, a non-parametric test is introduced, based on sampling from the Dirichlet distribution, which represents the null hypothesis of no association between the vectors. The p-value is calculated as the proportion of random comparisons in which the WSC exceeds the empirical value. The complete testing procedure, result interpretation, and numerical implementation are presented.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On the Statistical Significance of the Weight Similarity Coefficient in Multi-criteria Decision Analysis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wojciech Sałabun,
  • Jean Dezert,
  • Andrii Shekhovtsov,
  • Jarosław Wątróbski

摘要

This paper proposes a method for statistically assessing the similarity between two weight vectors, based on the Weights Similarity Coefficient (WSC). The WSC value measures the degree of similarity between two normalized vectors in the weight space. To determine the significance of this similarity, a non-parametric test is introduced, based on sampling from the Dirichlet distribution, which represents the null hypothesis of no association between the vectors. The p-value is calculated as the proportion of random comparisons in which the WSC exceeds the empirical value. The complete testing procedure, result interpretation, and numerical implementation are presented.