摘要
根据1952-2005年我国人口总量和GDP总量数据,建立变系数模型.采用逐步回归的方法来选择显著滞后变量子集,推导出系数函数的样条估计表达式,最后运用Bootstrap思想,进行点预测和区间预测。运算结果表明:对于该组数据,变系数模型能较理想地描述数据之间的内在结构,且具有较少的预测误差.
The varying-coefficient model is constructed with the data of population and GDP in 1952- 2005.First, choosing those significant variables by stepwise regressive approach, deducing the explicit polynomial spline estimation expression of coefficient function. Finally, giving the point forecast and interval forecast by bootstrap method. The result is that the varying-coefficient model can describe those data ideally with less of error.
出处
《数理统计与管理》
CSSCI
北大核心
2008年第5期755-759,共5页
Journal of Applied Statistics and Management
关键词
变系数模型
显著性变量
样条
varying-coefficient model, significant variables, spline
