Current browse context:
math.ST
Change to browse by:
References & Citations
Mathematics > Statistics Theory
Title: Confidence intervals with control of the sign error in low power settings
(Submitted on 31 Oct 2016 (this version), latest version 13 Nov 2019 (v3))
Abstract: When hypothesis tests of $H_0\theta=0$ have low power, it is possible that their rejection can frequently be accompanied by an estimate $\hat\theta$ that has the wrong sign and significantly exaggerated magnitude. Such sign errors are less likely when the confidence interval for $\theta$ is well separated from 0, as measured in units of the confidence interval's width. Sign errors can be controlled by declaring confidence in the sign of $\theta$ only when $H_0$ is also rejected at a smaller level $\alpha_2=2\alpha_1\alpha_S$, where $\alpha_S\le 1/2$ is a user specified upper bound on the probability of a sign error given that $H_0$ has been rejected. This procedure is a very simple form of inference after model selection: the selected model is that $\theta\ne0$ and the second inference is then on sign($\theta$).
Submission history
From: Art Owen [view email][v1] Mon, 31 Oct 2016 17:30:33 GMT (68kb,D)
[v2] Sat, 24 Aug 2019 00:20:29 GMT (79kb,D)
[v3] Wed, 13 Nov 2019 22:46:53 GMT (80kb,D)
Link back to: arXiv, form interface, contact.