On 2 Jun 2002 09:32:28 -0700, [EMAIL PROTECTED] (Awahab El-Naggar)
wrote:

> Dear Colleagues
> I have an experiment on 24 subjects. I get two measures for each subject in
> diffrent situations with 18 variables. I want to know if there are
> differences between the two situations in each of the 18 variables. I used
> SPSS paired t-test. Is it right for this experiment? What are other sutible
> alternatives?  Thanks for your advise.

I doubt if you really have 18 separate, distinct and independent
hypotheses.  So, you have a problem with 'multiple testing'  
which should be resolved.  

In the ordinary case that I see, a few variables are the important
ones.  They should be identified at the start:  the other variables
are secondary, subsidiary, exploratory, and otherwise, NOT in
the first rank.  

Among the ones that are thought to be the important ones, it
might be possible to set up a hierarchy, or create a composite,
or drop some variables -- based on correlations and whatever
other 'reliability'  information you may have.

If you end up with the single score, a t-test  is appropriate for the
overall hypothesis.  If you end up with 2 or 3 scores,  Hotelling's
T  is the basic version of the two-sample, multivariate test.
 - I'm not sure where you find a packaged test of that, but I
have not looked.


-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to