摘要
在变系数单指标模型的估计中,基于最小二乘惩罚函数方法是一种主流方法,一般情况下其具有良好的性质,但当参数分布非正态时,其估计结果不够稳健.在前人研究的基础上提出了一种基于Walsh平均的估计方法,并通过算例验证了估计的准确性及在参数分布非正态情形下估计的效率优势.
The method based on LS is mostly used when a varying-coefficient single-index model is applied. However,this way of estimation is not as much efficient as usual when parameters are non-normal distribution. To overcome this issue,Walsh-average-based spline estimation is proposed on the basis of previous studies. In addition, by using one example, the accuracy and efficiency of Walsh-average-based spline estimation under the non-normal distribution of parameters are verified.
出处
《内蒙古大学学报（自然科学版）》
CAS
北大核心
2017年第2期117-121,共5页
Journal of Inner Mongolia University：Natural Science Edition
基金
国家自然科学基金资助(71262022)
内蒙古自然科学基金资助(2014MS0708)
关键词
Walsh平均
变系数单指标模型
半参数模型
交叉验证
Walsh-average
varying-coefficient single-index model
semi-parametric model
crossvalidation
