Date:         Thu, 18 Sep 2008 13:15:41 -0500
Reply-To:     Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Subject:      Re: Log-linear model with "known" expected cell frequencies
Content-Type: text/plain; charset="us-ascii"

A slight twist is also of interest.  In the chi-square, fish are
replicates.  An alternative view would be that trials are replicates.  I
would feel more comfortable tackling the problem in this fashion, still
trying to test a single test of Habitat A vs B.  But again, I'm not sure
how to do so since my expected ratios are known, and inconsistent.  All
I can envision is 5 independent binomial tests, again combined under a
meta-analytic approach.
==========================================
I have figured a way to perform the latter.  I reduce the data to
proportions.  I then subtract the expected value based on the trial from
each observed.  If the observed are randomly distributed about the
expected, I should have a new random variate with mean=0.  Since my
expected values aren't near 0 or 1, the difference is likely close to
"normal".  So I can then use a t-test with 5 df.


Warren Schlechte
HOH Fisheries Science Center
5103 Junction Hwy
Mt. Home, TX 78058
Phone 830.866.3356 x.214
Fax    830.866.3549