home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!moe.ksu.ksu.edu!hobbes.physics.uiowa.edu!news.iastate.edu!iscsvax.uni.edu!sunfish!charlie.usd.edu!MGRANAAS
- Newsgroups: sci.math.stat
- Subject: testing for linearity
- Message-ID: <BzB43B.2zG@sunfish.usd.edu>
- From: mgranaas@charlie.usd.edu
- Date: Tue, 15 Dec 1992 15:23:34 GMT
- Reply-To: mgranaas@charlie.usd.edu
- Sender: news@sunfish.usd.edu
- Organization: The University of South Dakota Computer Science Dept.
- Nntp-Posting-Host: charlie
- Lines: 15
-
- The simplest test for linearity that I have found is to calculate the
- conditional means of y on x. Follow that with plotting and calculating
- the correlation coefficient. If your correlation is near +/- 1 your
- data are probably linear in form. If not, the plot should show a fairly
- clear picture of what shape the relation takes. If the plot does not
- show a fairly clear picture, then you do not have sufficient data at
- each level of x to dampen out the error. (This happens frequently, and
- at that point you're back to square one--sorry)
-
- Michael Granaas
- Dept. of Psychology
- University of South Dakota
- mgranaas@charlie.usd.edu
-
- How do I get a sig file into this post???
-