Article View: aus.stats.s
Article #260Re: [S] CIs for the conditional error rate of a classifier
From: Prof Brian D Ripley
Date: Fri, 03 Aug 2001 19:03
Date: Fri, 03 Aug 2001 19:03
26 lines
2374 bytes
2374 bytes
On Fri, 3 Aug 2001, Brian Steele wrote: > We are estimating the conditional error (or true error) > rate of a classification rule using n-fold (leave-one-out) > cross-validation using a relatively large data set (N=3000, > 15 classes). Does anyone have suggestions regarding how to > compute an approximate confidence interval for the > conditional error rate? I believe the CART book suggests using Poisson CIs, and I think they are fairly good. That is if you have R errors, assume R ~ Poisson(3000p) and derive the approximate CI for p. -- Brian D. Ripley, ripley@stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272860 (secr) Oxford OX1 3TG, UK Fax: +44 1865 272595 --------------------------------------------------------------------- This message was distributed by s-news@lists.biostat.wustl.edu. To unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with the BODY of the message: unsubscribe s-news ---
Technical Details
Message-ID:
<Pine.GSO.4.31.0108031950100.10150-100000@auk.stats>Path:
arc1-ca.pugleaf.net!archiver1.google.com!newsfeed.google.com!sn-xit-02!supernews.com!newsfeed.direct.ca!look.ca!news.maxwell.syr.edu!news1.optus.net.au!optus!news.usyd.edu.au!news.nsw.CSIRO.AU!cmis.csiro.au!owner-s-newsReferences:
<3B6AEA96.98C238C9@selway.umt.edu>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: fa931,66a8aca6bfab258d,start X-Google-Attributes: gidfa931,public X-Google-ArrivalTime: 2001-08-03 12:27:04 PST Newsgroups: aus.stats.s Path: arc1-ca.pugleaf.net!archiver1.google.com!newsfeed.google.com!sn-xit-02!supernews.com!newsfeed.direct.ca!look.ca!news.maxwell.syr.edu!news1.optus.net.au!optus!news.usyd.edu.au!news.nsw.CSIRO.AU!cmis.csiro.au!owner-s-news From: ripley@stats.ox.ac.uk (Prof Brian D Ripley) Subject: Re: [S] CIs for the conditional error rate of a classifier X-Nntp-Posting-Host: snowy.nsw.cmis.csiro.au Content-Type: TEXT/PLAIN; charset=US-ASCII Message-ID: <Pine.GSO.4.31.0108031950100.10150-100000@auk.stats> X-Cc: <s-news@wubios.wustl.edu> Sender: majordomo@mailhost.nsw.cmis.CSIRO.AU Organization: CSIRO Australia X-In-Reply-To: <3B6AEA96.98C238C9@selway.umt.edu> References: <3B6AEA96.98C238C9@selway.umt.edu> X-To: Brian Steele <bsteele@selway.umt.edu> Mime-Version: 1.0 Date: Fri, 3 Aug 2001 19:03:07 GMT Content-Length: 1114 X-Precedence: bulk X-Received: from max141.public.ox.ac.uk (max141.public.ox.ac.uk [192.76.27.141]) by toucan.stats.ox.ac.uk (8.9.0/8.9.0) with ESMTP id TAA25513; Fri, 3 Aug 2001 19:53:10 +0100 (BST) Lines: 26