潜类别分析原理及其在聚类分析中的应用

Principle of Latent Class Analysis and Their Application for Classification

目的探讨潜类别分析在多个二分类反应变量聚类分析中的应用。方法利用SAS软件ProcLCA模块,采用单一潜变量潜类别模型,对860名上海市公立性病门诊顾客有关性病预防知识的6个项目调查结果进行聚类分析。结果 4个潜类别的单一潜变量模型是数据拟合的理想模型(G2=38.52,df=36,P>0.05)。依据性病预防知识掌握情况,860名性病门诊顾客可以聚为4个亚群,每一亚群的人数分别为375人、208人、218人和59人。结论潜类别分析适用于多个二分类反应变量的聚类分析。

Objective To explore the application of latent class analysis for classification of multiple binary response variables.Methods Using latent class model,we gave 860 clients,who came from Shanghai public STD clinics a classification according to their knowledge about STD prevention.Results Four-class model was suitable to the data（G2=38.52,df=36,P〉0.05）,and the four sub-groups number of clients was 375,207,220,and 51 respectively.Conclusion Latent class model can resolve the classification of multiple binary response variables.