Dear Manon,
there is a parametric version suggested in 1959 by Bartholomew (Biometrika)
and the
more popular nonparametric version of Jonckheere. The second one is
generally treated in textbooks about nonparametrics, for the first one you
should
probably consult the original publication.
Furthermore the Cochran-Armitage trend test could also be of interest if
your classes
can be classified like 'doses'.
Which test you should use dpends strongly on your underlying distribution.
Kind regards
Robert
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|