Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: margins and xtmixed - revised
From
Ricardo Ovaldia <[email protected]>
To
[email protected]
Subject
Re: st: margins and xtmixed - revised
Date
Fri, 16 Mar 2012 09:15:33 -0700 (PDT)
Thank you Joerg. I have already plotted the trajectories using spaghetti plots and also after collapsing the groups. There is evidence of group effect differences and also the fixed effect of "group" in the model is significant.
I am more interested in the use of MARGINS do to post-hoc test.
Ricardo
Ricardo Ovaldia, MS
Statistician
Oklahoma City, OK
--- On Fri, 3/16/12, Joerg Luedicke <[email protected]> wrote:
> From: Joerg Luedicke <[email protected]>
> Subject: Re: st: margins and xtmixed - revised
> To: [email protected]
> Date: Friday, March 16, 2012, 9:11 AM
> You should plot the trajectories for
> each of your treatment groups.
> Then you can immediately see whether there are any
> differences across
> treatment arms or not. Whether these differences will be
> statistically
> significant will then simply be a matter of your sample
> size, study
> set-up, quality of measurements, observed/not observed
> covariates etc.
>
> However, if you really need a test, you want to have some
> kind of
> hypothesis. Your test is relatively unspecific: it tells you
> whether
> there is at least one difference between treatment arms at
> one given
> point in time.
>
> Anyway, I would say plot your trajectories and put
> confidence bands around them.
>
> J.
>
> On Fri, Mar 16, 2012 at 6:41 AM, Ricardo Ovaldia <[email protected]>
> wrote:
> > I am clarifying my prior query, because I messed up and
> perhaps confused folks with my prior post.
> > Here is the study:
> > Patients were randomly allocated to three treatment
> groups. Each patient was followed for a year and
> measurements made at baseline and every three months
> (month=0,3,6,9,12). Interest is on the effect of treatment
> overall, and at specify times (months) after controlling for
> sex and bmi. I used –xtmixed- allowing both a random
> intercept and random slope, and month by group interaction:
> >
> > .xtmixed ft i.group##month i.sex bmi|| id:month,
> cov(unstructured) mle
> >
> > To compare groups at given times, I used –margins,
> post- and –test-:
> > . margins, over(group month) post
> > . test 1.group#3.month=2.group#3.month=3.group#3.month
> > ( 1) 1bn.group#3bn.month - 2.group#3bn.month = 0
> > ( 2) 1bn.group#3bn.month - 3.group#3bn.month = 0
> >
> > chi2( 2)= 12.22
> > Prob chi2= 0.0022
> > So my questions are (1) most importantly is this
> correct? (2) are the delta-method SE being use by –test-?
> and (2) Is there a better way to compare groups at fixed
> times?
> >
> >
> > Thank you in advance,
> > Ricardo
> >
> >
> > Ricardo Ovaldia, MS
> > Statistician
> > Oklahoma City, OK
> >
> > *
> > * 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/