摘要
目的 探讨临床多变量类型资料 SAS聚类分析的方法。方法 采用 Jaccard系数进行样品间的相似性度量 ,构建非欧几里德距离矩阵 ,在 SAS数据步过程中完成上述计算 ,并生成分析数据集 ,通过调用 Cluster过程对资料进行聚类分析。结果 通过聚类分析 ,5 0例肝炎患者得到了合理归类 ,分类结果比较真实地反映了患者的临床特征。结论 利用 Jaccard系数度量样品间的相似性 ,构建距离矩阵的方法比较适用于临床多变量类型资料 ,操作简单 ,聚类效果也比较好。
Objective To investigate the method of using SAS cluster procedure to analyze the clinical data, which contains both continuous and scatter variables. Methods Jaccard coefficent is employed to measure the similarity of individuals, and the distance matrix is constructed. Cluster analysis can be doneintheSASdatastep. Results50 cases are classified reasonably. The results of cluster analysis reflect the patients clinical characteristics exactly. Conclusion The method is suitable for the clinical data that contains multiple variables. It can be operated easily and gets approving results.
出处
《中国医院统计》
2000年第2期69-71,共3页
Chinese Journal of Hospital Statistics
