Date:         Fri, 18 Oct 2002 17:20:17 +0200
Reply-To:     Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Subject:      Re: Q: How to get non-linear correlations?
Content-Type: text/plain; charset=us-ascii

Hi Jarod:

I Hope this helps:

"The Pearson product moment correlation coefficient r measures the degree of
linear association between X and Y. In the event of nonlinear but monotonic
relationships, a useful correlation coefficient is provided by the Spearman
correlation coeffcient r_s, which is a Pearson type correlation computed on the
ranks of X and Y, ranked separately..."

JD Jobson "Applied multivariate Data Analysis. Volume I: Regression and
Experimental Design"
pp 113. (Springer-Verlag, 1991)

Regards

Marta

(Sorry for the delay, my office was migrating from modem to cable-connection,
and I was without Internet connection for several days. Now, I'm finally
instaled in 21st century, at last!)

"Jarod_Russell, The_Pretender" ha escrito:

> Hi Marta,
>
> thanks fro the tip.
> Do you know some literature where this is written? Because in the literature
> I read this is not mentioned.
>
> BTW. I did found what I asked for - I can now draw the different 11
> trendlines with curvefit. :-)
>
> Do you or someone else know a way to make these for a lot of .sav-files?
>
> I wanted to check for the best R squread in each result and so make a
> decision.
>
> Jarod
>
> Marta wrote:
>
> Spearman correlation coefficient is NOT linear, it checks for any kind of
> correlation. You will get the same correlation value even if you try
> different transformations of both variables.
>
> HTH
>
> Marta