期刊
AUSTRALIAN & NEW ZEALAND JOURNAL OF STATISTICS
卷 46, 期 4, 页码 569-581出版社
WILEY
DOI: 10.1111/j.1467-842X.2004.00354.x
关键词
derivative estimation; non-parametric regression; simultaneous confidence band; SiZer; upcrossing probability
Curve estimates and surface estimates often contain features such as inclines, bumps or ridges which may signify an underlying structural mechanism. However, spurious features are also a common occurrence and it is important to identify those features that are statistically significant. A method has been developed recently for recognising feature significance based on the derivatives of the function estimate. It requires simultaneous confidence intervals and tests, which in turn require quantiles for the maximal deviation statistics. This paper reviews and compares various approximations to these quantiles. Applying upcrossing-probability theory to this problem yields better quantile approximations than the use of an independent blocks method.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据