Re: refrence for spearman partial correlation
- From: "Reef Fish" <Large_Nassau_Grouper@xxxxxxxxx>
- Date: 27 Apr 2006 15:46:54 -0700
Ray Koopman wrote:
Lee Sander wrote:
Hi,
using the same equation for partial correlation, i replace the
pearson correlation with spearman rank correlation.
What would be a correct reference to cite this.
thanks
les
Good question. Not sure what's a good reference to cite either.
Spearman correlations are Pearson correlations done on rank-
transformed data.
This much is correct and uncontroversial.
Since the Pearson computational formulas require
only that the variances be positive, and make no assumptions about
the form of the distributions, they are valid when applied to ranks.
This doesn't quite answer Lee's question, and is too much of a
simplification to the extent that it is misleading.
For the random variable r to have a useful interpretation for the
degree of linear relation, the distribution of r must be known or
at least approximated. There is much more to the question of
distribution of r, let alone Spearman and Spearman partial, than
it meets the eye.
In the Johnson, Kotz, Balakrishnan book, 2nd edition, on
"Continuous Univeriate Distributions", Volume 2, Chapter
32 is titled "Distribution of Correlation Coefficients", which
runs almost 100 pages, with 12 pages of references that
runs about 20 references per page.
If there is a good source for what you sought, I think you'll find
it there. :-)
-- Bob.
.
- References:
- refrence for spearman partial correlation
- From: Lee Sander
- Re: refrence for spearman partial correlation
- From: Ray Koopman
- refrence for spearman partial correlation
- Prev by Date: Re: Google's Statistical Report on sci.stat.math posting in April
- Next by Date: Re: Median of multi-dimensioned data
- Previous by thread: Re: refrence for spearman partial correlation
- Next by thread: decompose data
- Index(es):
Relevant Pages
|
|
