st: xtmixed estimation and postestimation

[Dennis Kramer <dkramerii@gmail.com>]

I am currently using the following syntax on an xtmixted function

--xi: xtmixed Test Outcome Teacher Evaluation Points  Student
Covariates School and Class Covariates || SCHOOL:|| TEACHER:
||STUDENT: , covariance(unstructured) var mle --

I am going to fit the model on a randomly generate portion of my
sample (90%) and restrict the other 10% to fit the models against.
However, I am unclear on the STATA commands to complete the following
portions:

    *estimate yhat =X’b for all observations  in the holdout

               I believe it is use fitdata -- fit the model -- use
"holdout" -- predict yhat

    * Get the residuals y – yhat for all observations in the holdout sample

    * Estimate a 2 level model on the residuals with teacher effect as
the only random effect

    * Retain the estimated teacher variance component

The remaining portions I am unclear on which commands will most
efficiently store and generate my estimates. Your help would be
greatly appreciated.

*
*   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/