变系数模型误差方差的估计

Error Variance Estimation of Varying Coefficient Models

变系数模型是由古典的线性模型发展而来,它们可以很好地检验函数系数随着协变量的变化程度.本文用PLR提出了变系数模型的误差方差的估计,并研究了它的渐近正态性,进一步用一个模拟例子来说明估计的结果是有效的.

Varying coefficient models are a useful extension of classical linear models. These models can easily examine how regression coefficient change over different groups characterized by certain covariates. In this article, the estimator of error variance by profile least-squares procedure is proposed and its asymptotic normality is studied. Furthermore, some simulations are conducted to examine the performance of our estimating approach and the results are substantial.