[SciPy-dev] Fast (O(n log(n)) ) implementation of Kendall Tau

Enzo Michelangeli enzomich@gmail....
Wed Sep 30 09:03:34 CDT 2009


Dear all,

A few days ago I posted to http://projects.scipy.org/scipy/ticket/999 a 
drop-in replacement for scipy.stats.kendalltau. My code implements the 
algorithm with complexity O(n log()) described by William R. Knight in a 
paper of 1966 archived at http://www.jstor.org/pss/2282833 , whereas the 
function currently in SciPy has complexity O(n^2), which makes it unusable 
with large data sets.

One potential issue here is that my code is in part derived from a Java 
implementation contained in the Java package called "Law" 
(http://law.dsi.unimi.it/software/law-1.4-src.tar.gz ).
That code is GPL'd, but its authors 
(http://law.dsi.unimi.it/index.php?option=com_contact&catid=4&Itemid=3 ) 
have already informally confirmed to me by e-mail that they have no 
objections to the release of my Python code under a BSD-style license. Now 
my question is: exactly which steps should I take in order to clear the path 
to a possible inclusion in SciPy?

All the best,

Enzo Michelangeli 



More information about the Scipy-dev mailing list