非线性计量经济模型的非参数估计方法研究

Study on Nonparametric Estimation of Nonlinear Econometric Model

本文提出使用核估计的方法构造平滑转移模型(STR)的非参数模拟最大似然估计(NPSML),给出了NPSML估计量的构造方法、渐近性质以及相应的核函数和窗宽的选择准则,并利用滑动窗宽算法对估计量的构造过程进行了改进。通过Monte Carlo实验证明,该方法是可靠的,并且当误差项存在序列相关时,此种估计量是稳健的。

This paper proposes nonparametric simulated maximum likelihood （NPSML） estimation for smooth transition model by kernel method. We provide a unified framework for constructing NPSML estimator and give rules for the choice of kernel function and bandwidth, and then use a slide variable bandwidth algorithm to improve its accuracy and reliability. Monte Carlo simulation study shows NPSML estimator is consistent and asymptotically efficient and robust when the error terms represent serial correlation.