home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!stanford.edu!bcm!convex!news.oc.com!eff!sol.ctr.columbia.edu!zaphod.mps.ohio-state.edu!sdd.hp.com!nigel.msen.com!emory!gatech!paladin.american.edu!auvm!UNLVM.UNL.EDU!BIOM030
- Message-ID: <EDSTAT-L%92110914314825@NCSUVM.CC.NCSU.EDU>
- Newsgroups: bit.listserv.edstat-l
- Date: Mon, 9 Nov 1992 13:26:38 CST
- Reply-To: April Milliken <BIOM030@UNLVM.UNL.EDU>
- Sender: "Statistics Education Discussion" <EDSTAT-L@NCSUVM.BITNET>
- From: April Milliken <BIOM030@UNLVM.UNL.EDU>
- Subject: Estimability?!?!?!?!
- Lines: 24
-
- I have a question that needs cleared up based on a student's question (I
- teach a SAS lab for an intro course for grads). I think this is more
- along the lines of linear models and I don't even know that I can phrase
- my question correctly.
-
- When deciding whether or not a hypothesis is estimable, say Ho: k`b=m ,
- is it usually the case that differences between parameter estimates will
- be estimable, but often times nothing else is (unbalanced case, maybe with
- missing data)? When working with generalized linear models, is it due to
- a non-unique generalized inverse...is this where the answer to estimability
- lies? That is, the solution vector is not unique, but the differences are;
- what else is unique (what else is estimable, and is it always true?)?
-
- I guess what I am looking for is guidelines for estimability. This all came
- about when my student made an error in writing his program, but then came up
- with "NON-ESTIMABLE" statements....extremely frightening to my novice
- programmers! If anyone can give me some guidelines or even ways to best
- explain this issue, I would be grateful!
-
- Thanks,
-
- April Milliken
-
- "I may be in Nebraska, but my heart is still at Kansas State University!"
-