re: st: rmanova or anova with repeated command, what to use?

["Airey, David C" <david.airey@vanderbilt.edu>]

.


        Baseline Placebo Active
        t0-t6    t0-t6   t0-t6
patient1
patient2
...
patient13


Some useful examples for using ANOVA in Stata, other than the manual, are
found at:

http://www.ats.ucla.edu/stat/stata/topics/anova.htm
http://www.ats.ucla.edu/stat/stata/faq/xtmixed.htm
http://www.stata.com/support/faqs/stat/anova2.html

The ANOVA command in Stata for two repeated measures factors is shown
in the last link here:

http://www.stata.com/support/faqs/stat/anova2.html#half713

You would use:

anova response subject time / subject#time group / subject#group time#group, repeated(time group)

You might not want to estimate a mean for every time point. 
You might instead want a linear or polynomial contrast. 
You might also want to use t0 to form differences with t1 to t6. 
Some people suggest using the baseline variable as a covariate too, 
even if you do that.

The sphericity assumptions of repeated measures ANOVA are probably 
not met in this experiment, so the p-values will be approximate 
after the G-G correction factor. You could alternatively try xtimixed 
with different covariance structure. A good book to help with that 
other than the manuals and the second link above is in the book store: 
http://www.stata.com/bookstore/lmm.html.

If you really wanted to look at just a time slice using ANOVA, 
you would have a simple repeated measures model with one within 
subject factor, e.g.,

anova response subject group if time == 1, rep(group)

or

bysort time: anova response subject group, rep(group)

-Dave



*
*   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/