Re: st: constraints insufficient for unique estimate

[Richard Williams <Richard.A.Williams.5@ND.edu>]

At 06:21 PM 8/10/2005 -0500, Zeynal Karaca wrote:
> When I run the regression it gives me the following error:
>
> constraints insufficient for unique estimate
> redundant or inconsistent constraints
>
> But, when I decrease the number of dummy variables for each category
> to 5, the regression is going through. But I need to use all the
> dummy
> variables becasue all of them seperated and unique and it is crucial
> for my dissertation. So, please help me out if I am missing something
> in the code or should I use a different definition for the
> constraint?
Does it run without the constraint?  I suspect not, but try to isolate the 
problem.

This is an awful lot of dummies - you are sure you have excluded at least 
one dummy for every categorical variable?  And that you are not spreading 
these way too thin, etc.? You may have to break down and combine some 
similar categories.

Personally, I suspect a perfect multicollinearity problem.  In an OLS 
regression vars would get dropped, but maybe cnsreg instead produces the 
error message you saw.

You could instead do this to impose your constraint:

gen x1x2 = x1 + x2

reg Y x1x2 x3... etc.

That way, if there is perfect multicollinearity, the reg routine will 
respond by dropping vars rather than giving you cryptic error messages.

You might also consider whether something like -areg- can meet your needs.

*
*   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/