Date:         Thu, 8 Jan 2004 10:32:02 -0500
Reply-To:     Magnus Alderling <magnus.alderling@SMD.SLL.SE>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Magnus Alderling <magnus.alderling@SMD.SLL.SE>
Subject:      Mixed modelling

Hi!

I have repeated measures and I'm planning to use the mixed procedure in SAS
to analyze my data.

We have salive cortisol levels which have been collected 4 times during 1
day. We have asked the individuals to give salive test:
1. Directly when they wake up
2. 1 hour after they woke up
3. At lunch-time
4. Just before they go to bed

As you can see the time-space between each measurement is not equal which
from what I understand the mixed procedure requires. But what I also have
noticed is except that the variance-covariance matrix called AR(1) can't be
used when the time space isn't equal, the variance-covariance matrices
called CS and UN in fact can be used. Before I start to estimate the fixed
effects, group and group*time I want to model the variance-covariance
matrix. Neither of the two groups that we have undergo any treatment.
That's why the curves for the the groups looks similar. Our hypothesis is
that the curve from one of the groups should be situated below the curve
from the other group. My question is if I should use the random or the
repeated statement in SAS? I know that the random statement model the
variance-covariance matrix for every group*subject and that the repeated
statement model the variance-covariance matrix for every
group*subject*time. Looking at the curves, how can I determine whether to
use the random or the repeated statement?

Best wishes!