纵向数据中分层线性模型的变量筛选

Principles and Methods According to Which HLM Models Deal with Longitudinal Data

10.6338/JDA.201002_5(1).0005

陈劲翔(Gin-Sheng Chen)；易丹辉(Dan-Huei Yi)

纵向数据 ； 分层线性模型 ； 变量筛选 ； Longitudinal Data ； Hierarchical Linear Model ； Variable Selection

Journal of Data Analysis

5卷1期（2010 / 02 / 01）

99 - 117

簡體中文

纵向数据一直都是生物医学统计探讨的热门话题，分层线性模型作为一种灵活的模型，在此领域得到了广泛的运用。然而，分层线性模型中层的结构虽然能够恰当地刻画纵向数据观测值嵌套于个体的特点，但同时也带来了一个问题──使模型中变量的数量大幅度增加，从而加大了分析和运算的难度。正因为此，分层线性模型要求建模者具有一定筛选变量的意识和能力。本文将基于分层线性模型处理纵向数据的原理和方法，讨论相关的变量筛选方法。

Longitudinal data has long been the popular topic in Biostatistics, and the Hierarchical Linear Model (HLM), as a flexible model, tends to be applied widely. Although the level structure of HLM helps describe the character of longitudinal data properly that observations are nested to individuals, a problem has been brought about at the same time and makes the model more difficult for analysis and calculation: the rapidly increasing number of variables in the model. Because of this, HLM model builders are required to possess the awareness and skills to filter proper variables. This paper will make a discussion on variable selection based on principles and methods according to which HLM models deal with longitudinal data.

1. Raudenbush, Stephen W.,Bryk, Anthony S.(2002).Hierarchical Linear Models: Application and Data Analysis Methods.Sage Publications.
2. Wang, SongGui、Yang, Hu(1994)。on the effect of centering and scaling on measures of multicollinearity in regression models。数理统计与应用概率，9(2)
3. 王济川、谢海义、姜宝法(2008)。分层统计分析模型—方法与应用。高等教育出版社。
4. 任红松、曹连莆、魏凌基、刘东旭(2003)。协方差分析的SAS实现方法。石河子大学学报，7(1)
5. 克雷伏特、里夫(2007)。多层次模型分析导论。重?大?出版社。
6. 易丹辉、李芃芃、童小军(2007)。中医临床疗效评价的统计方法。世界科学技术—中医药现代化，9(4)
7. 张岩波、何大卫、刘桂芬、王琳娜、郭明英(2001)。一般混合线性模型SAS的MIXED过程实现。中国卫生统计，18(4)
8. 张雷、雷雳、郭伯良(2003)。多层线性模型应用。教育科?出版社。
9. 盖笑松、张向葵(2005)。多层线性模型在纵向研究中的运用。心理科学，28(2)