home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!stanford.edu!bu.edu!olivea!netsys!ukma!nntp.msstate.edu!emory!europa.asd.contel.com!paladin.american.edu!auvm!VTVM1.BITNET!HARVEYRJ
- From: HARVEYRJ@VTVM1.BITNET (rj)
- Newsgroups: bit.listserv.stat-l
- Subject: standard error of r in small samples
- Message-ID: <STAT-L%92110518042128@VM1.MCGILL.CA>
- Date: 5 Nov 92 22:55:27 GMT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- Lines: 15
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
-
- I know this is a pretty simple question, but I have not found a simple answer
- yet! the question is, I need a formula for estimating the standard error of
- a correlation in a small sample (n = 20). I'd like to know what it would
- be both for the typical rho=0 case as well as for nonzero true rho. I've
- seen a bunch of approximations that work over N=50, but nothing as small
- as 20. the hard part is that there's very good reason to suspect that the
- observations on which the r is calculated are non-independent (i.e., a person
- can participate in rating 2-4 of the 20 "observations"). I'd be happy to
- just get an answer to the "pure" case w/o worrying about the complications
- caused by raters appearing more than once.
-
-
- r. j. harvey
- psychology
- virginia tech
-