非独立试验的组内相关与广义估计方程

INTRAUNIT CORRELATION OF CORRELATED OBSERVATIONS BASED ON GENERAL ESTIMATING EQUATIONS

目的 :探讨几种非独立试验资料的内部相关性。方法 :对非独立数据建立随机效应模型 ,利用广义估计方程之残差估计组内相关系数。结果 :组内相关反映了非独立试验资料间的共变关系或同趋性 ,由广义估计方程导出的组内相关是以方差齐性为前提的。结论 :对非独立试验 ,不仅要关心处理组间的效应 ,亦应关心非独立数据间的相关程度。用随机效应模型分析非独立资料 ,可得到更多的信息。

Objective: To estimate the intraunit correlation for correlated observations. Methods: Modeling random effect models for correlation observations. Estimating intraunit correlation by means of general estimating equations(GEEs). Results: Intraunit correlation is an efficiency statistics for estimating the correlation or the tendency to similarity for the observations in one cluster. The intraunit correlation derived from GEEs is under the conditions of homoscedasticity. Conclusion: For the trail including correlated observations, the focus is not only the effects of the treatments, but also the intraunit correlation. Modeling random effects models are most efficient for correlated observations.