用于数值分类的聚类分析

这里所谓聚类分析包括传统的聚类分析和作者提出的聚类统计检验分析。传统的聚类分析是常用的重要数值分类方法。但是，这个方法由于缺乏显著性检验，在国际上已长期公认是一个在多元统计中并不完整的方法。本文把聚类统计检验分析在其具有显著性检验的优越性上，和传统聚类分析作了对比分析。聚类统计检验分析区分有均匀聚类与中心聚类。前者用（１６）作为聚类判据，后者则用（２７），并且各用相应的公式（１７）－（２１）与（２

CLUSTER ANALISIS USED IN NUMERICAL CLASSIFICATION

The so-called cluster technigues in this paper consist of the traditional cluster analysis and the cluster analysis with statistical test, the latter being suggested by the author. Traditional cluster analysis is a practically important technique often used in numerical classification. Howeyer, for a long time this technique has been known as being incomplete in multivariate analysis for lack of tests of significance. In this paper the cluster analysis With statistica1 test is illustrated for its excellence of test of significance in comparison with the traditional cluster technique.Homogeneous clustering and central clustering are distinquished in this cluster analysis with statistical test. The cluster criterion of the former and of the latter can be obtained by Eq. (16) and Eq. (27)respectively in connection with the corresponding formulas (17)-(21) and (22)-(26). Central clustering is sometimes appropriate for large-scale numerical classifications, while homogeneous cluster for both small-scale classifications and classifications in experiments.