If variables in a regression model are collinear, there are an infinite
number of parameter estimates (i.e. "coefficients") that will give the
best fit. Stata has a default procedure that "removes" variables from
the model (i.e. by constraining their coefficients to be zero) to make
the solution unique. However in simple linear regression, you could
impose your own linear constaints such as (1) b1 +3*b4 - b7 = 23, 2(2)
13b3-b5=-0.1234, etc. using -cnsreg-. Another is the nocons option
(constraining b0 = 0). If you impose your own constraints, there must be
enough of these to make the system linearly independent, otherwise Stata
will still set some coefficients to zero.
However if you are doing something other than simple linear regression,
such as -glm- , Stata may not allow the specification of linear
constraints in the same manner that -cnsreg- does.
Al Feiveson
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Silvia
Mendolia
Sent: Sunday, March 25, 2007 11:41 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: Collinear variables
Hello everybody,
does anybody know how to tell to Stata that collinear variables do not
have to be removed?
I know there is an option collinear, but it doesn't seem to work for
linear or probit regressions.
I am aware of the possibles negative consequences of this, but I just
need to know how to do this...
Thank you very much,
Silvia
*
* 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/
*
* 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/