Mann Whitney test.

[John Darrington]

Jason and I were discussing the Mann-Whitney test on IRC last week,
which inspired me to look up the literature on it.

One interesting paper I found is Bergmann, Ludbrook and Spooren,
"Different Outcomes of the Wilcoxon-Mann-Whitney Test from Different
Statistics Packages", The American Statistician 54(1), 2000, pp 72-77

From their conclusion:
 "... statistics packages provide very inadequate documentation in
 their manuals... As a consequence, the results can be dangerously
 misleading.
 Investigators cannot rely on the popular, general purpose,
 microcomputer statistics programs ... to provide an accurate outcome
 of the WMW test."

Earlier in the paper, they say the following about spss:

 "The second `exact' p value presumably results from Dineen and
 Blakeksley (1973) algorithm, even though this is invalid when there
 are ties."


We ought to think very carefully therefore when implementing this in
PSPP, and I rather agree with the authors that most stats programs'
documentation is non-existent.

J'


--
PGP Public key ID: 1024D/2DE827B3
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.




_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[Jason Stover-3]

On Wed, Sep 24, 2008 at 01:37:49PM +0800, John Darrington wrote:

> Jason and I were discussing the Mann-Whitney test on IRC last week,
> which inspired me to look up the literature on it.
>
> One interesting paper I found is Bergmann, Ludbrook and Spooren,
> "Different Outcomes of the Wilcoxon-Mann-Whitney Test from Different
> Statistics Packages", The American Statistician 54(1), 2000, pp 72-77
>
> From their conclusion:
>  "... statistics packages provide very inadequate documentation in
>  their manuals... As a consequence, the results can be dangerously
>  misleading.
>  Investigators cannot rely on the popular, general purpose,
>  microcomputer statistics programs ... to provide an accurate outcome
>  of the WMW test."
>
> Earlier in the paper, they say the following about spss:
>
>  "The second `exact' p value presumably results from Dineen and
>  Blakeksley (1973) algorithm, even though this is invalid when there
>  are ties."
>
>
> We ought to think very carefully therefore when implementing this in
> PSPP, and I rather agree with the authors that most stats programs'
> documentation is non-existent.

Then I think we should pick a "good" algorithm from the literature and
implement that. Then mention it in the manual, as well as the reason
for the discrepancy with the other software.

-Jason


_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev