On Sun, 2 Jun 2002 05:47:54 -0500
"Jerry Harder" <[EMAIL PROTECTED]> wrote:

> 
> "Yvette" <[EMAIL PROTECTED]> wrote in message
> [EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> > Hello all,
> >
> > 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 MR AR NS <S
> > Cash .9968(<.0001) -.2374(.0016) -.0074(.9216) -.9722(<.0001)
> .0453(.5679) -.0579(.4650)
> > Earn .4588(<.0001) -.2444(.0011) -.0039(.9585) -.9796(<.0001)
> .0429(.5889) -.0670(.3983)
> > FR -.0609(.4231) -.0334(.6606) .2313(.0019)
> .3997(<.0001) -.1177(.1356) -.0415(.6000)
> > MR -.1018(.1798) .0178(.8147) .1330(.0767)
> .0461(.5423) -.0416(.5932) -.1235(.1118)
> > AR -.0225(.7672) -.0081(.9143) .9643(<.0001) .0319(.6726) .0637(.4207)
> .0640(.4180)
> > NS .1663(.0349) .0840(.2892) -.1386(.0785) -.0167(.8301) -.1425(.0704)
> .1263(.1016)
> > <S -.0482(.5429) -.0399(.6145) -.0391(.6211) -.1888(.0145) .0040(.9597)
> .0810(.2949)
> >
> >
> > Thanks
> 
> Keep in mind that Pearson and Spearman test two different hypotheses: Person
> tests a linear relationship between the variables and Spearman tests a
> monotonic relationship. Since Spearman uses ranks it is less susceptible to
> outliers and "influential" values. For those variable pairs where there is a
> an apparent big difference, run some regression diagnostics on both the
> original values and ranks. This should give you a better understanding of
> the differences.
> 
> --
> Good luck,
> 
> Jerry Harder

That is well put.  As a linear relationship is a special case of a monotonic 
relationship, there is little reason to use Pearson's r if what you want is a P-value. 
 And when Pearson's r is optimum, Spearman's rho is still 0.96 efficient.

Frank Harrell


> remove spamnein from address to reply
> 
> 


-- 
Frank E Harrell Jr              Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine  http://hesweb1.med.virginia.edu/biostat
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to