变系数混合模型的光滑样条推断

为了拟合纵向数据和其他相关数据,本文提出了变系数混合效应模型(VCMM).该模型运用变系数线性部分来表示协变量对响应变量的影响,而用随机效应来描述纵向数据组内的相关性, 因此,该模型允许协变量和响应变量之间存在十分灵活的泛函关系.文中运用光滑样条来估计均值部分的系数函数,而用限制最大似然的方法同时估计出光滑参数和方差成分,我们还得到了所提估计的计算方法.大量的模拟研究表明对于具有各种协方差结构的变系数混合效应模型,运用本文所提出的方法都能够十分有效地估计出模型中的系数函数和方差成分.

Inference in Varying-Coefficient Mixed Modelsby Using Smoothing Spline

Varying-coefficients mixed model (VCMM) isproposed for longitudinal data and the other correlated data. Thismodel allows flexible functional dependence of the response variableon the covariates by using varying-coefficients linear part topresent the covariates effects, while accounting the within-subjectcorrelation by using random effect. In this article, the coefficientfunctions are estimated by using smoothing spline and restrictedmaximum likelihood is used to estimate the smoothing parameters andthe variance components simultaneously. The performance of theproposed method is evaluated though some simulation studies, whichshow that both the coefficient functions and variance componentscould be estimated well for the VCMMs with all kinds of covariancestructures.