Ok, Fran, a lot of thanks! 
 
But in any case I think that it would be better to implement a menu in the window of the restricted model, because sometimes the models are very complicated (think, for example, about the case I am working with: a system of cost functions, with about 10 restrictions among different parameters) and I don’t see you suggestion very operative for that cases.
 
Javier
 
If you make the substitutions
 
Y3 = Y- X3
X13 = X1-X3
X23 = X2-X3
 
\
and run the regression of Y3 on X13, X23 and other variables you get
confidence intervals for the coefficients on X1 and X2.  You can then
estimate the coeficient on X3 as 1-X1-X2 and can estimate its standard error
manually.  Alternatively you could make a second transformation
 
Y1 = Y - X1
X21 = X2 - X1
X31 = X3-X1
 
and run a similar regression and get the same answers as previously but
without any calculations
 
Please forgive this notation which is using for coefficients and variables.
I hope that it does not cause confusion
 
Regards
 
John
 
2011/5/30 Javier García <javigarcia83 at yahoo.es>
 
> The problem is that I have more than one restriction (and not very easy to
> implement). Besides, if I have for example X1, X2 and X3 and I want to
> introduce the restriction X1+X2+X3=1 then, how I can recover the
> coeffcient/s that doesn't appear in the restricted model??? I would have to
> obtain them "manually"...
> 
> Best
> Javi