Re: st: Re: Re: Repeated Measures ANOVA: contrasts

["Joseph Coveney" <stajc2@gmail.com>]

Vincent Koppelmans wrote:

. . . The 7 time points I have can be grouped as follows:
Time point 1 and 2: pre treatment
Time point 3, 4, and 5: treatment
Time point 6 and 7: post treatment

Hence, I do not expect a difference between TP 1 and 2, but I do expect some 
cumulative effect from time point 3 to 5.

This is why I wanted to use contrasts instead of calculating averages per 
condition (i.e., pre, during, and post treatment).

What would be a good and valid solution here?

--------------------------------------------------------------------------------

I don't have anything to add to David's excellent answers to your numbered 
questions.  My only suggestion for this last one would be:  if you're doing 
formal hypothesis testing, then stick to your study's written protocol in order
to avoid getting stung by researcher degrees of freedom; otherwise, for 
exploratory work, I'd favor a graphical approach that is guided by the science.

By the way, those contrasts *are* calculating averages per condition.  The 
practical difference between a contrast (1/2 1/2 -1/3 -1/3 -1/3 0 0) after ANOVA
and averaging before a t-test (1/2 1/2) - (1/3 1/3 1/3) = 0 is in the test 
statistics' denominators.  As the simulation illustrated, the latter is less 
susceptible to (pretty strong) autocorrelation.

Joseph Coveney

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/