[Scipy-tickets] [SciPy] #1484: "Kulsinski" dissimilarity seems wrong

SciPy Trac scipy-tickets@scipy....
Tue Aug 16 00:08:54 CDT 2011


#1484: "Kulsinski" dissimilarity seems wrong
---------------------------+------------------------------------------------
 Reporter:  muellner       |       Owner:  peridot    
     Type:  defect         |      Status:  new        
 Priority:  normal         |   Milestone:  Unscheduled
Component:  scipy.spatial  |     Version:  0.9.0      
 Keywords:                 |  
---------------------------+------------------------------------------------

Comment(by muellner):

 The formula in the second link which warren.weckesser gave (it's called
 Kulczynski-Cody index there) agrees with my two links. This formula has
 the advantage that it agrees with R, so it would not confuse users.

 But I agree with warren.weckesser that a definitive reference is
 desirable. Unfortunately, it's in Polish and from 1928. Let's follow back:

 Henning and Hausdorf (doi: 10.1080/10635150500481523) cite Shi
 (doi:10.1016/0031-0182(93)90084-V). This author mentions two similarity
 (sic) measures, both attributed to Kulczynski. The reference is

 Kulczynski, S., 1928, Zespoly róslin w Pieninach. Bull. Int. Acad. Pol.
 Sci. Lettres, Sér. B, Suppl., 2:57-203

 I haven't read the Polish paper. But according to Shi's paper, the
 measures are

 "Kulczynksi unnamed 1" = 1 - (dissimilarity measure from warren.weckessers
 first link)

 "Kulczynksi unnamed 2" = 1 - (dissimilarity measure from the other three
 links)

 I vote for the second version, since more people appear to use this one.
 warren.weckesser already stated it: 1 - 0.5 * ntt * (1.0/u.sum() +
 1.0/v.sum())

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1484#comment:4>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list