Re: st: coefficients constraint in nlsur and nl does not work

[vwiggins@stata.com (Vince Wiggins, StataCorp)]

Mariko <mariko.wt@gmail.com> has written that -nlsur- does not do what
he wants when he specifies the same linear combination twice.  Using a
simple but silly example,

    . nlsur (y = {xb: x1 x2} + {xb: x1 x2})

estimates 4 parameters rather than the 2 Mariko wants.  We agree with
Mariko. Estimating 2 parameters is much more useful in this case.  We
will change this behavior in a future update.  

In the interim, Mariko can follow Brian Poi's <brian@poiholdings.com>
final suggestion and explicitly enter the second expression using the
parameter names that the first linear combination creates, e.g.,

    . nlsur (y = {xb: x1 x2} + {xb_x1}*x1 + {xb_x2}*x2)

Or, he can dispense with linear combinations altogether and name his
two parameters whatever he wishes.

    . nlsur (y = {b1}*x1 + {b2}*x2 + {b1}*x1 + {b2}*x2)

I believe the reason that Mariko is the first to encounter this
misfeature is because it does not occur if the linear combinations
occur strictly in two different equations.  For example,

    . nlsur (y1 = {xb: x1 x2})  
            (y2 = {xb: x1 x2})

estimates just two parameters.

 
-- Vince 
   vwiggins@stata.com

 
*
*   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/