Rich Ulrich <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>... > On 1 Jun 2002 15:56:46 -0700, [EMAIL PROTECTED] (Yvette) wrote: > > > I conducted a pearson (upper diagonal)/spearman (lower diagonal) > > correlation test in SAS. However, the results differ significantly > > between the two tests. Can anyone offer any suggestions as to why > > these results would differ so much. > > > > RESULTS OF CORRELATIONS: > > > > Cash Earn FR AR > > Cash .997 -.23(.002) -.97(.000) > > Earn .45(.000) -.24(.001) -.98(.000) > > FR -.06(.4) -.03(.6) .40(.000) > > AR -.02(.7) -.01(.9) .96(.000) > > - Here are the problem-correlations without the tab-characters > and extra digits, and extra variables. It has already been > answered (extreme Pearson r determined by the outliers). > I had to edit it in order to read it, and I thought other people > might be interested in the original numbers, too. > > The Kendall rank-correlation is less responsive to the > extreme *rank* matches than the Spearman is. But the > differences above tell you, already, you want to look at plots > or raw numbers.
Outliers could explain the Cash/Earn Cash/AR and Earn/AR figures, however, the FR/AR figures (where the Spearman is much stronger than the Pearson) might indicate an almost monotonic but nonlinear relationship. Or, it could be caused by a couple of outliers on the same side, not both on a straight line with the rest of the data. Look at the data and you'll probably see what's causing it. Are any of these variables ratios? Glen . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
