半参数时变系数模型的序列相关检验

Testing Serial Correlation in Semiparametric Time-varying Coefficient Model

本文提出了两个统计量来检验半参数时变系数模型的序列相关:一个用于检验半变系数时序模型的有限阶序列相关,另一个用于检验半变系数平行数据模型的有限阶序列相关.在误差过程为鞅差的零假设下,所提出的两个检验统计量服从渐近正态或卡方分布.蒙特卡罗模拟研究表明所提出的检验统计量具有良好的有限样本性质.

This paper proposes two test statistics for testing zero finite-order serial corre- lation in semiparametric time-varying coefficients model： one is for semivarying-coefficient time series model, and the other is for semivarying-eoefficient panel data model. Under the null assumption of a martingale difference error process, the proposed test statistics are shown to have the asymptotic normal or X2 distributions. Simulation studies show that the proposed test statistics have good finite sample properties.