st: Hierarchical Linear Modelling in Survival Anaysis

[Lee Panas <lpanas@brandeis.edu>]

I am trying to do hierarchical linear modeling in survival analysis and 
am wondering whether this is possible in STATA or whether I should try 
another option.

I have a sample of about 8,000 patients served in about 100 substance 
abuse facilities.  The dependent variable is time to arrest (TIMEARR) 
from start of treatment.

The primary independent variable is Engagement or ENGAGE.  ENGAGE=1 if 
the patient received three services in the first 45 days of treatment. 
ENGAGE=0 otherwise.  Co-variates include prior criminal justice 
involvement (PRIORCJ 0/1), FEMALE (0/1) and about 20 others. Those who 
were arrested within 45 days from the start of treatment are dropped 
from the analysis.

I want to control for clustering within facilty and to also estimate the 
effect of facility-level variables such as facility size (FACSIZE) and 
facility engagement rate (PCTENGAGE).  I have controlled for clustering 
using the SHARED command:

stset timearr, failure(arrested)
stcox engage priorcj female , shared(facility)

Is there a way to also estimate the effect of FACSIZE and PCTENGAGE.  I 
could put them in the model as individual level variables, but facility 
level variables are preferred.

Thanks for any advice which you can provide.

Lee






*
*   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/