David / Maarten,
Thank you for your assistance. I have a supplementary question. Following anova I often test and adjust for multiple comparisons. With your example:
. test, test(m1) mtest(sidak)
( 1) a[1] - a[2] = 0
( 2) a[1] - a[3] = 0
( 3) a[1] - a[4] = 0
( 4) a[2] - a[3] = 0
( 5) a[2] - a[4] = 0
( 6) a[3] - a[4] = 0
Constraint 3 dropped
Constraint 4 dropped
Constraint 5 dropped
---------------------------------------
| F(df,21) df p
-------+-------------------------------
(1) | 0.71 1 0.9571 #
(2) | 4.45 1 0.2513 #
(3) | 30.08 1 0.0001 #
(4) | 1.60 1 0.7740 #
(5) | 21.53 1 0.0008 #
(6) | 11.39 1 0.0170 #
-------+-------------------------------
all | 11.63 3 0.0001
---------------------------------------
# Sidak adjusted p-values
Is this possible to do after xtmixed as there is no design matrix?
Thanks.
Paul
--- On Wed, 31/12/08, David Airey <[email protected]> wrote:
> From: David Airey <[email protected]>
> Subject: Re: st: Re: XTMixed / Repeated Measures
> To: [email protected]
> Date: Wednesday, 31 December, 2008, 3:44 PM
> use http://www.ats.ucla.edu/stat/stata/examples/kirk/rb4,
> clear
> anova y a s, repeated(a)
> test _coef[a[3]] = _coef[a[4]]
> anova, regress // note a[4] was dropped
> xi: xtmixed y i.a || s: // note a[1] was dropped
> test _Ia_3 = _Ia_4
> display r(chi2)/r(df) // scale to show F the same, but here
> df 1
>
> ...shows the same results, yes, and wrt your question, the
> coefficients shown are relative to the dropped group, in
> this model. You can also change which group gets dropped.
>
>
>
>
>
> On Dec 31, 2008, at 8:37 AM, Paul Bransom wrote:
>
> > Sorry - forgot the subject on my earlier posting
> >
> > Hello,
> > I am trying to teach myself how to use xtmixed for
> repeated measures anova. Using anova:
> >
> > anova m method id,repeated(method)
> > I have a significant effect of method and I can
> compare the methods:
> > test _coef[method[1]] = _coef[method[2]] ( 1)
> method[1] - method[2] = 0
> > F( 1, 27) = 58.27
> > Prob > F = 0.0000
> >
> > However when I use xtmixed
> > xi:xtmixed m i.method || id:
> >
> > -----
> >
> >
> ------------------------------------------------------------------------------
> > m | Coef. Std. Err. z
> P>|z| [95% Conf. Interval]
> >
> -------------+----------------------------------------------------------
> > -------------+------
> > _Imethod_2 | -12.6 1.650589 -7.63 0.000
> -15.8351 -9.364905
> > _Imethod_3 | -15.1 1.650589 -9.15 0.000
> -18.3351 -11.8649
> > _Imethod_4 | -11.1 1.650589 -6.72 0.000
> -14.3351 -7.864905
> > _cons | 38.6 2.253023 17.13 0.000
> 34.18416 43.01584
> >
> > I can again compare all the methods:
> >
> > . test _Imethod_2 = _Imethod_3
> >
> > with the exception of method 1 as _Imethod_ is the
> "reference" in the equation.
> > My question is - how can I compare the effect of
> method 1 to the other methods or is the fact that these
> coefficients have a significant p value sufficient (or am I
> missing something that is so obvious)?
> >
> > I am using Stata 10.
> > Thanks,
> > Paul
> >
> >
> >
> >
> >
> > *
> > * 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/
>
> *
> * 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/
*
* 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/
