On 16 Dec 2002 19:04:07 -0600, Brian Sandle
<[EMAIL PROTECTED]> wrote:

> In talk.politics.drugs Rich Ulrich <[EMAIL PROTECTED]> wrote:
[snip ]
> 
> Perhaps you can point me to the derivation of how the long Pearson 
> recipe reduces to the much shorter Spearman when ranks are entered 
> rather than scores. 
[ snip, rest]

JEH Shaw has posted (in sci.stat.edu), pointing to the derivation.

Jim Clark has posted, and I hereby thank him for answering
so completely the questions (directed to me) about the
fundamentals of statistical vocabulary, looking at data 
with transformations, etc.

Here is a reference in the literature that inspired me to focus
on 'rank-transformation'  as the alternative to saying
'nonparametric.'   Conover's textbook is also a standard.

 - Conover, W.J., and R.L. Iman. 1981. Rank Transformations 
as a bridge between parametric and nonparametric statistics. 
The American Statistician 35:124-133.


> 
> Now if you wish to do partial rank with significance I can supply a VAX 
> program based on Siegel and Castellan. I am not sure how reliable the 
> p-levels. 

We statisticians tend to be conservative -- in the direction
of *not*  wanting to use a procedure that is guaranteed to
give 'significant'  p-levels  owing to artifact.  I sure hope 
that S&C  have packaged sufficient warning along with that.

As it should be clear by now:  anyone who wants to do
a partial-correlation of a Spearman  rho can rank-score the
variables, and feed them into the corresponding programs
for a Pearson correlation.   - So long as the correlations are
rather small, this should be reliable:  as estimates of raw
correlation or partial correlation, and of p-level.  
(Complication:  What is 'rather small'?)

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
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