Date:         Sun, 30 Mar 2008 12:24:21 -0700
Reply-To:     Pierre Nouvellet <pierrenouvellet@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Pierre Nouvellet <pierrenouvellet@HOTMAIL.COM>
Organization: http://groups.google.com
Subject:      mixed model + generalized linear model
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hi, I'm no expert on stats, and i have found a bit of a challenge
with
this problem:
I was first trying to deal with it with SPSS, but I've been advised to
use either SAS or R
I understand the 2 languages are rather similar...

So,
I have 4 variables:
Ra: count, number of road traffic accident involving an animal
Y: year
M: month
T: temperature for the month

my data are in long format.
i want to test: Ra~Y+M+M*T .
more specifically i want to know if monthly temperature influence the
monthly count, and if so which month have a significant impact (slope
different from 0!).

since each year a different effort is put to record the accident I
need to include it. also I am happy to assume that in reality the
count should be the same each year... So I just want to correct the
yearly effort...
since each month may have different count i include it too.

what i really want is to know the SLOPES of the relation for each
month (controling for effect of year and month) and if it is
significantly different from zero!

I understand year should be taken as random factor? (1) (I am not
interested in knowing the influence of year, since I assume the
variation only due to effort).
and if you use random factor, better to include them in a mixed
model...

As fixed factor, I include M and interaction M*T    (2)   ,

As random factor, I include year, any tips on covariance type and
contrast I should use?(3)

so it is really 3 question in one... As well as have some input to
wether the choice of model is appropriate...

thanks for taking the time....