Date:         Wed, 25 Feb 2009 11:20:27 -0500
Reply-To:     Ryan Utz <rutz@AL.UMCES.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ryan Utz <rutz@AL.UMCES.EDU>
Subject:      Permutations-getting started with zero experience

Hi all,

I'm embarking on a permutation procedure that I have no experience with in
SAS. In the end, what I'd like to do is run permutations of an ANCOVA model
with subsets of a particular dataset and get a distribution of model
parameters (p-values, etc.). I hope this is possible. Here is how the data
are constructed:

data one; input site_ID bin $ urban flow wshed_nest $; cards;
1 A .4 13 1
2 A .7 15 2
3 A 2.5 19 3
4 A 12 29 4
5 A 15 28 5
6 A 22 30 6
7 A 28 31 6
8 A 24 27 6
9 B .5 20 7
10 B .7 21 8
11 B 1.1 25 9
12 B 5.5 33 10
13 B 10 39 11
14 B 15 44 12
15 B 17 50 12
16 B 16.5 48 12
;

(the dataset is much larger than this-this is just an example).

What I would like to do is run the model:

proc glm data=one; class bin; model flow=bin urban bin*urban; run;

But notice the variable 'wshed_nest'. What I need to do is randomly take
just one site from a group of sites where this variable is identical (for
instance, just one from where the site is '12' and one from where it is '6',
etc.), include all sites where the wshed_nest is unique, run the ANCOVA, do
it again and again with a different random combination, and summarize
certain ANCOVA model parameters somewhere (inside or outside of SAS-either
works).

I'm not looking for a final solution from this group-just where to start.
PROC PLAN looks like a lead, but since I've never done anything like this
before, I wanted to see what the world thought first. Any suggestions, tips,
PROC recommendations...?

Thanks,
Ryan