Date: Tue, 29 Nov 2005 10:35:22 -0500
Reply-To: Peter Flom <flom@NDRI.ORG>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <flom@NDRI.ORG>
Subject: Re: PROC CORR filtering out non-significant correlations from the
Content-Type: text/plain; charset=US-ASCII
>>> Sameer Patil <sameer@CRCE.ALUMNI.NET> 11/29/2005 9:21 am >>> wrote
Thanks for the quick response.
Actually, I am working with survey data. I am doing exploratory
analysis to find interesting relationships amongst variables. Part of
my goal is to discover relationships (if any) between all these
factors, and looking at statistically significant correlations seemed
to be the first step to start with.
I am not a stats expert, so I am not sure I follow the reasoning
behind why I would get more than 5000 significant correlations.
Well, you said you had 310 variables. That leads to a LOT of correlations. Indeed,
it leads to 47,895 of them.
Now, suppose all those variables were just random noises. Well, just by chance,
some of them would be significantly related to each other. About 10% at a p < .1;
about 5% at p < .05. So, that's not the way to go.
310 is a LOT of variables. What are they? (not a whole list....but what sorts of things)?
You will probably want to do some sort of data reduction before proceding.