变系数混合模型的光滑样条推断(英文)

Inference in Varying-Coefficient Mixed Models by Using Smoothing Spline

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.

Varying-coefficients mixed model （VCMM） is proposed for longitudinal data and the other correlated data. This model allows flexible functional dependence of the response variable on the covariates by using varying-coefficients linear part to present the covariates effects, while accounting the within-subject correlation by using random effect. In this article, the coefficient functions are estimated by using smoothing spline and restricted maximum likelihood is used to estimate the smoothing parameters and the variance components simultaneously. The performance of the proposed method is evaluated though some simulation studies, which show that both the coefficient functions and variance components could be estimated well for the VCMMs with all kinds of covariance structures.