Re: st: question about hierarchical model

[Joerg Luedicke <joerg.luedicke@gmail.com>]

On Thu, Apr 21, 2011 at 4:59 PM, Visintainer, Paul
<Paul.Visintainer@baystatehealth.org> wrote:
> I'm stumped on how to approach this model.
>
> Suppose I have 30 teachers each of whom have a single performance rating from the principle (or headmaster).  Each teacher has 20-30 students (not crossed) who have completed student evaluations.  The question is what is the association between performance ratings and student evaluations?
>
> What I'm confused about is that I have level-2 dependent variable with level-1 and level-2 predictors (e.g. teacher age and teacher gender)
>
> I thought that the following model might work:
>
> .xtmixed t_perform s_eval t_age t_sex || s_eval: , mle
>
>
> But I don't think this is correct model, because it regress Yi on Xij and Xi predictors?
>
> Can this problem be addressed using -xtmixed- or is a different approach required?
>

You are right regarding your concerns about your outcome measure. You
cannot have a level2 outcome in a 2level model since you would
basically predict a constant within each group (teachers in your
case). Regarding your research question:

"what is the association between performance ratings and student evaluations?"

I would start with simply taking the average of student ratings for
each teacher and then correlate the 2 measures or run a regression
with those 30 teachers. That could answer your question already.

But others may have better ideas.

J.

*
*   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/