Re: st: Re: Interpretation of xtmixed repeated measures

[Philip Ender <ender97@gmail.com>]

Dear Janet,

I think the -anovalator- program (findit anovalator) will do what you
want.  -anovalator- uses -margins- for its computations. After, the
-xtmixed- command enter this:

anovalator group, main fratio

This will compute the 'overall' main effect and scale the chi-square
as an F-ratio/

Now for the caveats.
1) Although -anovalator- reports an F-ratio, it does not give p-values
for F.  It only gives the p-value associated with the chi-square.
Neither -margins- nor -anovalator- are aware of denominator degrees of
freedom.  In large samples the difference between the chi-square and F
p-values can be ignored.  However, in small samples the differences
are bigger.
2) If your group#time interaction is significant then trying interpret
the main effect for group becomes difficult.  A significant
interaction implies that the group effect is different at different
times.  You can use -anovalator- to coupute the omnibus two-way
interaction and tests of simple main effects as follows:

anovalator group time, 2way simpe fratio

I hope this is helpful


Phil
-- 
Phil Ender
UCLA Statistical Consulting Group

Janet Hill  wrote:


 I am trying to learn to use xtmixed for repeated measures designs. In
the example from the Repeated Measures FAQ:

anova lhist group / dog|group time time#group if dog !=6,repeated(time)
There is a significant effect of group. However when I use:
xtmixed lhist group##time if dog !=6 || dog:
margins group##time, post asbalanced
     |            Delta-method
             |     Margin   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       group |
          1  |  -2.093812   .3719332    -5.63   0.000    -2.822788   -1.364837
          2  |  -2.552977   .4294714    -5.94   0.000    -3.394725   -1.711228
          3  |  -.8932172   .3719332    -2.40   0.016    -1.622193   -.1642416
          4  |  -2.479522   .3719332    -6.67   0.000    -3.208498   -1.750546
             |
etc.

I can compare individual groups but I cannot see how to get an
'overall' effect as in the anova.
I would be grateful for some advice on how to achieve this or a
reference to a tutorial on xtmixed and margins.

Thanks,
Janet
*
*   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/