|Date: ||Fri, 16 Aug 1996 07:28:08 +0000|
|Reply-To: ||Edel Christoph MSM PH DE <christoph.edel@DEWE.MHS.CIBA.COM>|
|Sender: ||"SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>|
|From: ||Edel Christoph MSM PH DE <christoph.edel@DEWE.MHS.CIBA.COM>|
|Subject: ||Re: Incomplete Blocks & Testing Hypothes|
you might need to be more specific to receive a comprehensive answer. Are
you referring to
Type I, II, III, or IV tests?
If you consider Type III hypotheses, which are of the form L*b=0, then for
the same set of coefficients L, a re-ordered parameter vector b will result
in a new hypothesis. Re-ordering the components in the MODEL statement
yields such a different parameter vector, since
the design matrix X (or in the case of GLM, X'X) is constructed according to
a set of rules
which can be found on p.928 ff of your SAS STAT manual (Version 6, Fourth
As far as I understand, Type II hypotheses should not be affected, but if I
am wrong, please will someone correct me.
>>>>>>>>>>>>>>> original posting: >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Date: Thu, 15 Aug 1996 20:51:54 +0000
From: pg <cristian@TAMBAQUI.ESALQ.USP.BR>
Subject: "Incomplete Blocks & Testing Hypothesis."
Hi, SAS-L users:
I am looking for some information (books, articles, suggestions)
about why the hypothesis tested are different if you change the
order of the components in the model statement in PROC GLM,
dealing with incomplete blocks designs. I know is something
related with Searle's "R notation" and I have the paper, but I need
further reading.Comments, sugestions bienvenidas.
Thanks in advance
Cristian A. Carranza | Fortuna plango vulnera, stillantibus
Cristian@tambaqui.esalq.usp.br | quod sua michi munera subtrahit rebellis
| Verum est, quod legitur, fronte capillata,
| sed plerumque sequitur, ocassio calvata.