Date:         Tue, 29 Mar 2005 11:07:16 -0800
Reply-To:     cassell.david@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV>
Subject:      Re: Correlation between MAT Score v/s GPA
Comments: To: srihari swamynathan <srihari1981@rediffmail.com>
In-Reply-To:  <20050329013037.29050.qmail@webmail29.rediffmail.com>
Content-type: text/plain; charset=US-ASCII

srihari swamynathan <srihari1981@rediffmail.com> replied:
> On Tue, 29 Mar 2005 David L.Cassell wrote :
> >I see that Wensui Liu has already given you some excellent advice.
> >
> >Depending on what your hypothesis really is, PROC CORR may be the
> >answer, and it may not.  What do you mean?  Here are three different
> >options you might be contemplating.  Note that they are VERY
different!
> >
> >1  "MAT and GPA rise together linearly."
> >2  "MAT and GPA rise together, but not necessarily linearly."
> >3  "If MAT goes up, then GPA goes up too."
> >
> >Option 1 is a linear correlation.  This looks like PROC CORR with
> >Pearson correlation.
> >
> >Option 2 is a monotonic relationship.  As Wensui noted, now you're
> >into the realm of Spearman correlations.
> >
> >Option 3 is not correlation at all, but causality.  You CANNOT
conclude
> >this using correlations.  In fact, you cannot attribute causality
using
> >only statistics, no matter what.
> >
> >So again I ask, what do you mean?  What is your underlying
hypothesis?
>
> well, my hypothesis is to show that the MAT score has a significant
> effect on GPA. In simpler terms, one who is scoring good marks in
> MAT has a higher GPA than who doesnt have.

I'm sorry, but you need to quantify that a bit more.  Can you state your
hypothesis numerically?  Statistically?  Are you assuming causality or
not?
Your response is still too vague for me to tell *exactly* what you mean.

David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician