Re: st: Re: new to xtmixed - basic question

[Ricardo Ovaldia <ovaldia@yahoo.com>]

Thank you Clyde,

In my model I did have -treatment- as a categorical variable, I just inadvertently dropped the "i." when I posted it to Statalist. When I use  - i.treatment##c.time##c.time2- in the model all interaction terms are not significant and neither is the treatment effect. When I do not interact treatment and time, treatment is significant.

Ricardo

Ricardo Ovaldia, MS
Statistician 
Oklahoma City, OK


--- On Thu, 7/21/11, Clyde Schechter <clyde.schechter@einstein.yu.edu> wrote:

> From: Clyde Schechter <clyde.schechter@einstein.yu.edu>
> Subject: st: Re: new to xtmixed - basic question
> To: statalist@hsphsun2.harvard.edu
> Date: Thursday, July 21, 2011, 9:30 AM
> In your first model:
> 
> The model is of the form:
>  xtmixed met treatment x1 x2 x3 || participantid:,
> covar(unstructured)
> 
> Stata would _not_ know what the time variable is and no
> time effects would
> be modeled.
> 
> Your second model:
> 
> . gen time2=time^2
> 
> . xtmixed met treatment x1 x2 x3 time time2||
> participantid:,
> covar(unstructured)
> 
> gets you closer.  But here, the only thing you detect
> about treatment is
> whether or not the trajectories of met over time are
> vertically displaced
> from each other in the three treatment groups. 
> (Actually, you won't even
> get that because specifying treatment the way you did,
> Stata will assume
> you want treatment treated as a 1 df numeric variable
> rather than as a
> 3-level factor variable.)  But I would imagine you
> would like to consider
> the possibility that the treatment leads to trajectories
> with different
> steepness and curvature.  So I think what you really
> want is something
> like:
> 
> xtmixed met i.treatment##time##time2 x1 x2 x3 ||
> participantid:,
> covar(unstructured).
> 
> The above will work if your time variable is simply
> ordinal.  If it is
> actually an elapsed time you may want to treat it as
> continuous, in which
> case you will want to substitute
> i.treatment##c.time##c.time2 in the
> above.
> 
> (I am assuming that you are not interested in looking at
> treatment effect
> modification by the covariates.)
> 
> Hope this helps.
> 
> 
> Clyde Schechter
> Department of Family & Social Medicine
> Albert Einstein College of Medicine
> Bronx, NY, USA
> 
> Please note new e-mail address: clyde.schechter@einstein.yu.edu
> 
> *
> *   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/