Re: st: Inequality constraints in a regression (new question)

[Steven Samuels <sjsamuels@gmail.com>]

--

Herve-

This is exactly what I predicted in an earlier post, and is a  
consequence of the geometry of least squares. It occurs when the  
constraint is a4>=a2, but the original least square estimates have  
a2<a4.   The LS solution with a4 = a2 has a lower residual sum of  
squares than any solution with a4>a2; that's what makes it  
"constrained least squares".  You can get the constrained common  
coefficient directly by entering into  -reg- the single covariate  z  
=(g_ca + g_ca_t).  Or use -nl- with the term   {a2_4}*(g_ca + g_ca_t).

Steve

Steven J. Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783

On Nov 14, 2010, at 4:15 AM, Herve STOLOWY wrote:

Dear All:

Last week, I asked a question relating to inequality constraints in a
regression. I received several interesting answers who suggested to use
-nl- for the constraint: a2 < a4 and to replace a4 by  ({a2}+
exp({c}))*g_ca_t

Here is the new model.

nl (VTA = {a1}*g_t + {a2}*g_ca + {a3}*g_i + ({a2}+ exp({c}))*g_ca_t +
{a5}*g_ca_i + {a6}*g_t_i + {a7}*g_ca_t_i), initial(a1 1 a2 1 a3 1 c 1 a5
1 a6 1 a7 1)

To recover the value of a4, I use -nlcom-:

nlcom a4 : _b[a2:_cons] + exp(_b[c:_cons])

My problem is the following:

- in the first nl regression, I get a2 =  2.110647
- in the nlcom command, I get a4 = 2.110647. (I precise that all the
elements of the line seem to be identical: same standard error, same t,
same p, same confidence interval).

I am confused because I thought that this solution would make a4 higher
than a2, which is an important constraint to me. Here, a4 = a2.

Best regards

Hervé Stolowy


***********************************************************
Professeur/Professor
HEC Paris
Departement Comptabilite Controle de gestion / Dept of Accounting and
Management Control
1, rue de la Liberation
78351 - Jouy-en-Josas
France
Tel: +33 1 39 67 94 42 - Fax: +33 1 39 67 70 86
mail: stolowy at hec dot fr
web: http://www.hec.fr/stolowy


*
*   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/


*
*   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/