home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!UCLAMVS.BITNET!IARSPAL
- Message-ID: <EDSTAT-L%92122303032637@NCSUVM.CC.NCSU.EDU>
- Newsgroups: bit.listserv.edstat-l
- Date: Wed, 23 Dec 1992 00:02:00 PST
- Reply-To: "Peter A. Lachenbruch" <IARSPAL@UCLAMVS.BITNET>
- Sender: Statistics Education Discussion <EDSTAT-L@NCSUVM.BITNET>
- From: "Peter A. Lachenbruch" <IARSPAL@UCLAMVS.BITNET>
- Subject: andy taylor's problem
- Lines: 10
-
- I'm no expert in latin squares, but a general linear models approach might work
- here. You have 3 d.f. for treatment, 3 d.f. for rows, 3 d.f. for columns,
- leaving 6 d.f. for error. You might create a covariate which is 1 if the in-
- sect pest was present and 0 otherwise. This will use up 1 d.f. and will pick
- out the effects you want. In general, the Tukey single d.f. for non-additivity
- is sometimes suggested - it's found in most anova texts - I think Snedecor and
- Cochran has it - but I don't know if many computer programs offer it (BMDP migh
- t be a good place to start).
- Good luck and Merry Christmas.
- Tony Lachenbruch
-