On Thu, 09 Mar 2000 20:20:47 GMT, [EMAIL PROTECTED] wrote:
< ... >
> 1 Independent Variable (Group)
> 13 Dependent Variables (various IQ and cognitive testing scores)
>
> The MANOVA was significant and I wanted to perform post hoc tests on
> Group. Here then is the caviat:
>
> Group N
>
> 1 148
> 2 5051
> 3 134
> 4 69
> 5 103
>
> As you can see the differences in the various N's is quite extreme.
>
> My question is to seek advice on what would be the appropriate post hoc
> test to use with unequal N's and especially when they are this
> extreme. Some have suggested Tukey HSD for unequal N, Hochberg's GT2,
> and others. Any help would be greatly appreciated.
You can see from my stats-FAQ that I don't care very much for followup
tests. I guess, I like them even less when you try to adapt them to a
multivariate circumstance.
What is it, that you hope to conclude? Somehow, I find it hard to
believe that all your hypotheses ought to be completely symmetrical,
when you have one group that is 70 times as big as another. (However,
only *one* group is huge, so testing across unequal Ns is not as
distorted as it would be if two groups were huge.)
Say, you also do have *13* variables ... so you say.
Five groups with 13 variables -- that gives you a discriminant
function, if you want to put the common name on the analysis. Is
there just one root that separates the groups? Or, is there a second
dimension to worry about?
If there is just one, then the 5 are located on a number-line, and the
*FULL* information about testing is provided by giving the equivalent
of the ordinary t-tests between each of the pairs, if you can figure a
reasonable way to do that. If these variables make up one narrow
dimension, I would do a a-priori scoring of the dimension, and do my
testing with ANOVA, and t-tests; otherwise, it does get difficult.
Anyway. The full information about "effect" is given by citing the
multivariate means, so anyone can see the actual distances.
--
RichUlrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
