Hi.
I'm running a model with many many dummies. The purpose of the analysis is
to find the relation between access to care and cost. In my program I
compute predictive margins for those access variables and then I run a
simulation to compute confidence intervals for the predictive margins.
Everything works with the exception of the simulation. The simulation code
per se is working but not with all the dummy variables I have in the model.
It works when I control using 60 dummies for age/sex, race, and mental
health comorbidities. But I also need to control for general medical
comorbidities and for these I am using a risk adjustment tool that has 174
variables. When I enter these in the model (together with the other 60 plus
the access variables) the simulation part fails as follows:
. corr2data $indv constant, n($number) means(betas) cov(varcovar) clear
matrix not positive definite
r(506);
In the regression command, all variables entered in the model are retained
but could I still be having colinearity problems? If so, is there any
command or program that I could use to figure out which variables are
collinear? Looking at a correlation matrix of all these variables would be a
nightmare.
Any comments or suggestions are greatly appreciated.
Thanks, Maria
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
