Thieret N L, Anderson D H
J Clin Psychol. 1985 Mar;41(2):197-202. doi: 10.1002/1097-4679(198503)41:2<197::aid-jclp2270410211>3.0.co;2-t.
Cluster analysis and the MMPI were used together to determine whether the heterogeneous diagnostic groups could be identified in an inpatient psychiatric population. The effectiveness of four clustering techniques in separating diagnostic groups was compared using percent of coverage, between-sample replication, between-method replication, cluster differences on age, and cluster differences on diagnosis. The results showed cluster analysis to be a poor diagnostic discriminator when compared to diagnostic groups, which were considered the "true" clusters of the sample. The final conclusion of the study was that cluster analysis should not be used to form groups that would be diagnostically labeled because each analytic technique produced misleading results.
聚类分析和明尼苏达多相人格测验（MMPI）被共同用于确定在住院精神科患者群体中是否能够识别出异质性诊断组。使用覆盖百分比、样本间复制、方法间复制、年龄上的聚类差异以及诊断上的聚类差异，比较了四种聚类技术在区分诊断组方面的有效性。结果表明，与被视为样本“真实”聚类的诊断组相比，聚类分析是一种较差的诊断区分方法。该研究的最终结论是，聚类分析不应被用于形成将被诊断标记的组，因为每种分析技术都会产生误导性结果。
