【提 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。

    

【Abstract】 Objective toexplore the relationship between the variance component model and the linear regressionmodels in fitting 2 level hierarchically-structured data.Methods bycomparative analysis for the parameter estimates between three linear regression mdels and2 level variance component moddel.Results the parameter estimates andit's stability of linear regression models is associated with the variation of explanatoryvariable between and within level 2units,and the relationship of parameter estimates andit's stability between linear regression models and variance component model is associatedwith level 2 residual variance or intatra-unit correlation.Conclusion thethree linear regression models are inappropriate whten data are hierarchicallystructured,and both the single random error term of linear regression model can bedecomposed into corresponding levels of hierarchical structure,and the randowm effects oflevel 2 units can be estimated by variance component model.

     ......