[Scipy-tickets] [SciPy] #865: Kendalltau divides by zero
SciPy
scipy-tickets@scipy....
Wed Feb 4 03:06:49 CST 2009
#865: Kendalltau divides by zero
-------------------------+--------------------------------------------------
Reporter: icmurray | Owner: somebody
Type: defect | Status: new
Priority: normal | Milestone: StatisticsCleanup
Component: scipy.stats | Version: devel
Severity: normal | Keywords: kendalltau cleanup stats divide by zero nan
-------------------------+--------------------------------------------------
It may be that I have not understood the definition of the kendall-tau
properly, but I think there is a problem with the kendalltau when one or
both input lists ranks all elements as being equal. In particular:
{{{
>>> from scipy.stats import kendalltau
>>> kendalltau([1,1,1],[3,3,3])
(nan, nan)
>>> kendalltau([1,1,1],[3,3,0])
(nan, nan)
}}}
Now, according to the wikipedia entry for kendall tau:
http://en.wikipedia.org/wiki/Kendall%27s_tau
I'd expect:
{{{
>>> kendalltau([1,1,1],[3,3,3])[0] = 0
}}}
Since the number of concordant pairs is 6, the number of discordant pairs
is 6, and the total number of possible pairs is 6. i.e. (6 - 6)/6 = 0
--
Ticket URL: <http://scipy.org/scipy/scipy/ticket/865>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.
More information about the Scipy-tickets
mailing list