[R] spearman rank correlation problem

Robert W. Baer, Ph.D. rbaer at atsu.edu
Tue Mar 16 15:43:44 CET 2004


> Martin Maechler writes:
> It seems that this needs to be documented, since you are right,
> the    1 - 6 S(d^2) / (n^3 - n)    ( d = r1 - r2 ; r{12} := rank(x{12}) )
> formula is also in use as *definition* for Spearman's rank correlation.

Since this topic is here...

Although I seen the above formula in a couple of books, but I have never
seen a derivation.  Does anyone have a reference where it is derived
(apologies to math and statistics types for whom this must seem child's
play)?  Examining the derivation is probably the best way to understand the
effect of ties when using William Morgan's function for Spearman.

Rob Baer




More information about the R-help mailing list