非参数核密度估计在异方差模型中的应用

Application of Heteroscedastic Models in Non-parametric Kernel Density Estimation

针对非参数核密度估计中最优窗宽选择在实际建模中的不足,提出新的最优窗宽选择的迭代方法,克服使用传统的经验法则所带来的局限性。并在此基础上用新的非参数核密度估计ML方法研究中国股票市场,通过与极大似然估计对比论证此方法的有效性和可行性。实证分析表明,通过与实际值的模拟对比,运用非参数估计技术得到上证指数日收益率的拟合值要优于极大似然估计的拟合值。

Since the heteroscedastic model is inadequacy in practical modeling, a new estimating method is put forward, namely non-parameter kernel density estimation-ML method. A new method of computing window width is proved, which overcomes the shortage of customary computing. Using this method, we do some researches on the Chinese stock market, and find that it gives a better feasibility and effectiveness, comparing with the maximum likelihood estimation. Empirical analysis also shows that the nonparametric estimates technique is superior to the maximum likelihood estimation techniques.