纵向数据模型的稳健估计及经验似然推断

A Robust Estimation and Empirical Likelihood Inference for Longitudinal Data Models

本文针对纵向数据模型,利用指数平方损失函数构建了一种稳健的估计方法,并通过广义估计方程将数据间的相关性考虑到估计中,提高了估计的效率,推导出稳健的经验似然比函数,由此得到参数估计以及进行统计推断.通过蒙特卡洛模拟发现,无论假定的工作相关矩阵是否准确,本文构建的估计都具有较强的稳健性,且用本文方法进行统计推断能够得到更加稳健的结果。

By use of the exponential squared loss function, this paper proposes a robust estimation for longitudinal data models. First we assume the working correlation matrix and construct robust generalized estimation equations. Then we deduce the robust empirical likelihood ratio function for estimation and inference. The Monte Carlo simulation shows that the proposed estimator is robust as well as the proposed statistical inference even though the working correlation matrix is misspecified.