固定设计点情形下回归函数值的区间估计
Pointwise Interval Estimation of the Regression Function for Fix-Designed Nonparametric Regression Model
-
摘要: 对于固定设计点情形下的非参数回归模型,基于核估计的方法给出了回归函数的逐点置信区间和修正偏差置信区间. 所给置信区间在相对较弱的条件下是渐近精确的. 考虑到窗宽的选取对核估计和区间估计的重要影响,给出了适合所给置信区间的核估计窗宽的迭代选取方法.模拟结果显示迭代法收敛,且选取的窗宽表现良好,置信区间有令人满意的覆盖率.Abstract: In this paper, the pointwise confidence interval and bias-corrected confidence interval are developed based on kernel estimation. Considering that the bandwidth plays an important role in kernel estimation and interval estimation, an iterative method is proposed to select proper bandwidth in estimation. The estimated intervals are proven to be asymptotically accurate and their satisfying performace are shown by simulation results.
下载: