摘要
目的：为了在当前的有序样本聚类法〔１～３〕中消除在多变量情形下变量量纲对聚类结果的影响和获得对形式结果的准确解释。方法：从原方法的基本思想出发，考虑对资料先作离均差平方和单位化的变量预处理，并修补了对聚类结果的一般解释方式———即视一般的多点类为变化相对平稳“段”，而单点类为脉冲跳跃点或构成“过渡类”。结果：对文献〔１～３〕的资料的重新分析和对药物麻醉前后脑电图的若干指标的分段分析，均取得了符合客观实际的结果。结论：我们的修正是准确、合理。
bjective:In order to eliminate the influence of dimensions of variables upon the cluster results in the situation of multiple var iables and obtain the accurate interpretation for the formal results in the ordered sample cluster analysis at present〔1～3〕.Methods:According to the basic idea of the original method,the variable pretreatment which tranforms the sum of squares of deviations of data into a unit value is first considered,and some repairments are made for the general interpretation pattern of the results in which general multipointsclasses are regarded as relatively stable classes,and singlepointclasses as pulse points or component parts of transitionclasses.Results:Reanalysing data from〔1～3〕and making the classification analysis for some variables of EEG during anaesthesia,the results are well consistent with the objective reality.Conclusion:Our correction is accurate,reasonable,and effective
出处
《中国卫生统计》
CSCD
北大核心
1998年第2期21-24,共4页
Chinese Journal of Health Statistics
