摘要
目的探讨对于具有两个水平层次结构的数据拟合方差成份模型与线性回归模型的关系。方法通过三类传统线性回归模型与两水平方差成份模型参数估计的对比，分析其内在联系。结果线性回归模型参数估计以及估计的稳定性与自变量在水平２单位间和水平２单位内的变异大小有关，线性因归模型与方差成分模型参数估计及其稳定性的关系与水平２残差方差或单位内相关系数大小有关。结论当数据具有层次结构特征时，三类线性回归模型均存在问题，方差成分模型可将线性回归模型的单一随机误差项分解到与数据层次结构相对应的各水平上，并估计水平２单位的随机效应。
Objective to explore the relationship between the variance component model and the linear regression models in fitting 2 level hierarchically structured data.Methods by comparative analysis for the parameter estimates between three linear regression mdels and 2 level variance component moddel.Results the parameter estimates and it's stability of linear regression models is associated with the variation of explanatory variable between and within level 2units,and the relationship of parameter estimates and it's stability between linear regression models and variance component model is associated with level 2 residual variance or intatra unit correlation.Conclusion the three linear regression models are inappropriate whten data are hierarchically structured,and both the single random error term of linear regression model can be decomposed into corresponding levels of hierarchical structure,and the randowm effects of level 2 units can be estimated by variance component model.
出处
《中国卫生统计》
CSCD
北大核心
1999年第1期14-16,共3页
Chinese Journal of Health Statistics
关键词
层次结构数据
方差成分模型
线性回归模型
hierachically structured data variance component model linear regression models intra unit correlation
