分类与回归树——一种适用于临床研究的统计分析方法

介绍分类与回归树(classification and r egression trees, CART)的发展历史、结构、组成和特点.CART包括分类树和回归树两部分,分类树的结果变量是分类变量,回归树的结果变量是连续变量.CART是一种树型结构,由树结和连线组成, 在末端的树结又称为终止结.CART可分析同质性较差的数据,采用替代变量的方法解决缺失数据问题,不要求数据的分布,可同时利用各种类型的数据.CART的树型结构与临床思维十分接近,有利于CART在临床研究中的应用. CART可用于临床研究数据分析,其应用范围有待于不断扩展.

Classification and regression trees (a statistical method suitable for clinical researches)

To introduce classification and regression trees (CART). The development, structure, main elements and features of CART were introduced. CART was structured by two parts, classification tree and regression tree. Classification tree used nominal variable as outcome, and regression tree used continuous variable as outcome. Tree structure was the feature of CART, and it was made up of tree notes and lines. The terminal tree notes were named end notes. CART was suitable for non homogeneous data analysis, using surrogate to replace missing data, suitable for any distributed data, and all kind of variables. The tree structure of CART was very like clinical thoughtway and suitable to explain results for clinical purpose. CART is a new statistical method suitable for clinical data analysis. The applied range of CART in clinical researches needs to be expanded.