[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: RE: Constrained regression

From   "Nick Cox" <>
To   <>
Subject   st: RE: RE: Constrained regression
Date   Mon, 11 Aug 2008 18:56:26 +0100

Better answer:

Also, you just have one fewer parameter than you thought, e.g. b1, b2, 1 - b1 - b2, not b1, b2, b3. 

-----Original Message-----
From: [] On Behalf Of Nick Cox
Sent: 11 August 2008 18:11
Subject: st: RE: Constrained regression

Constraints in Stata mean linear constraints. Inequalities don't qualify. 

As I understand it your coefficients can't be arbitrarily positive or zero and yet still sum to 1. Each must also be between 0 and 1. 
That last is sometimes met by reparameterising in terms of the logit. Nevertheless, the question of implementing the last constraint still remains. 


Höchle, Daniel 

I would like to estimate regression model

y_t = a1*x1_t + a2*x2_t + ... ak*xk_t + e_t

with the following constraints:

a1 + a2 + ... + ak = 1

I tried to use -cnsreg- and -nl- but somehow did not manage estimate the model. 

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index