Limit theorems for kernel-type estimators for the time of change

被引:4
|
作者
Grabovsky, I
Horváth, L
Husková, M
机构
[1] Univ Utah, Dept Math, Salt Lake City, UT 84112 USA
[2] Charles Univ Prague, Dept Stat, Prague 18600 8, Czech Republic
关键词
change-point; kernel; Brownian motion; strong approximation; limit distribution;
D O I
10.1016/S0378-3758(00)00100-2
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
We use kernel-type estimators to estimate the time of change in the mean in a sequence of independent observations. Assuming that the size of the change is small two types of limit distributions are derived. The forms of the limit distributions depend on the behavior of the kernel at the end points. The argmax of a two-sided Brownian motion with polynomial drift is a possible limit, while the normal distribution is the limit when the kernel is zero at both boundaries. (C) 2000 Elsevier Science B.V. All rights reserved.
引用
收藏
页码:25 / 56
页数:32
相关论文
共 50 条