Please don't trim previous posts so that readers can't even see what
you're responding to
and have to check back.
Just because there is scope to model error structure doesn't mean it is
compulsory or necessarily
a good idea. I agree with you to that extent. But Mann-Whitney U gives
you no scope at all to do
that. Sure, it can be a useful simple test.
But with modelling commands e.g.
glm mpg foreign
glm mpg foreign, f(gamma)
glm mpg foreign, link(log)
I can start exploring which assumptions matter and which don't (not to
mention other aspects such
as assumptions for standard errors).
[email protected]
modelling error structure with a pilot study in pharmacology sounds
funny,
specially with 6 mice in each group, Nick. Math and statistics sometimes
blind researchers.
I would not try to answer anything with a pilot study (in
pharmacology/biochemistry, etc..).
Pilots are done to see the direction of the effect. If it shows a
signal,
we keep going. If not, we stop and try other alternatives. Mann-whitney
sounds ok, at least in my modest view.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
