Jen Zhen wrote:
Many thanks for your replies, the Hotelling test seems to do a good
job for what I want.
One follow-up question though would be whether it might be even better
to instead use (for both the continuous and the categorical, but
ordered variable) a non-parametric test, like the
Wilcoxon-Mann-Whitney test? Power is not an issue, as our sample is
quite large...
--------------------------------------------------------------------------------
He can correct me if I'm mistaken, but I believe Austin's point was to test
everything once in one fell swoop rather than in a piecewise fashion. With the
Wilcoxon-Mann-Whitney test (-ranksum-, see my earlier post in this thread),
you'd still be testing one response variable at a time. Along Austin's
suggestion, you can use -gllamm-, which can accommodate mixed response types
(continuous, ordered-categorical) in a single multivariate model, or, if you'd
like to stick with a nonparametric test, then take a look at -mv2snp-, which is
a user-written multivariate nonparametric two-sample test that you can download
and install from SSC (-net stb 42 snp14-).
Joseph Coveney
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
