Dear list,
I've encountered the following strange behaviour when computing a quantile
of Wilcoxons rank sum distribution using qwilcox():
> qwilcox( 0.95, 36, 36)
Problem in cut.default(x, ..1): breaks must be given in ascending order
and contain no NA's
Use traceback() to see the call stack
Similar sample sizes (even or odd, and at least the ones, I've tried)
don't seem to present any problem, e. g.:
> qwilcox( 0.95, 34, 34)
[1] 1307
> qwilcox( 0.95, 35, 35)
[1] 1383
> qwilcox( 0.95, 37, 37)
[1] 1540
> qwilcox( 0.95, 38, 38)
[1] 1621
I've checked the archive, but didn't find anything. Does anybody have an
idea about what is happening above, and even better: how to work around
this problem (bug?) ? Thank you.
Best regards -- Gerrit
PS: S-PLUS Version 6.0 Release 1 for Sun SPARC, SunOS 5.6 : 2000.
-------------------------------------------------------------------------
Dr. Gerrit Eichner Mathematical Institute of the
gerrit.eichner@math.uni-giessen.de Justus-Liebig-Univ. Giessen
Tel: +49-(0)641-99-32104 Arndtstr. 2, 35392 Giessen, Germany
Fax: +49-(0)641-99-32029 http://www.math.uni-giessen.de/Stochastik
-------------------------------------------------------------------------
|