[SciPy-user] help with scipy.stats.mannwhitneyu

josef.pktd@gmai... josef.pktd@gmai...
Thu Feb 5 11:46:48 CST 2009


On Thu, Feb 5, 2009 at 12:29 PM, Sturla Molden <sturla@molden.no> wrote:
> On 2/5/2009 6:23 PM, josef.pktd@gmail.com wrote:
>
>> you mean that it uses bigu for the p-value calculation but reports
>> smallu as the test-statistic?
>
> Yes
>

Given that it works, I didn't want to spend time on this, but wikipedia again:

     "therefore, the absolute value of the z statistic calculated will
be same whichever value of U is used."


As I understand it, because the sum U1+U2 is fixed (given the sample
sizes), many properties are equivalent, i.e.
U1 - meanU = - (U2 - meanU)

so whether bigU or smallU is used in the calculation of z doesn't
matter, I have no idea why in this specific implementation both are
calculated if smallU would be enough.

Josef


More information about the SciPy-user mailing list