Prev: oracle missing expression error (Sql)
Next: interpreting interaction term in survival model
From: rjs on 10 Dec 2009 07:00
Hi all

Just wondering whether anyone knows of a workaround to get the id
level coefficients ouit of proc mdc when using the mixedlogit syntax?

Tech support and documentation say there's no inbuilt option to do
this.

Thinking it might be possible to feed the initial regression a range
of sufficiently uncorrelated inputs with a missing dependent
variable , then do a secondary regression on the forecast data to
derive the id level coefficients.

Seems cumbersome though- have about 30000 individuals and hundreds of
choice events per individual - any thoughts?

Would be most grateful for suggestions - alternative software looks
expensive..

Code example below:


proc mdc data=discchoice outest=c;
model decision = var1 var2 var3/
type=mixedlogit
nchoice=3
optmethod=qn
covest=hess

mixed=(normalparm=var3);

id pid;
output out=probdata pred=p xbeta = a;
ods output ParameterEstimates = a;

Kind regards

Robert
 | 
Pages: 1
Prev: oracle missing expression error (Sql)
Next: interpreting interaction term in survival model