How to impose convexity constraint to regression coeffs ?

[vge]

Hello,

Assume one has a regression of the kind:
DEPVAR = B1*Y1 + B2*Y2 + errors

which is run in RATS as follows:
linreg(noprint) DEPVAR start end
# Y1 Y2

Now, I want to impose a constraint on B1 and B2 so that B1+B2 is equal or less than 1 ?
How do I do that ??????

Many thanks for your help !
V.

[TomDoan]

If that's all you're trying to do, it's very simple:

1. Run the original regression. Check the sum of the coefficients. If it's <= 1, you're done.
2. If it's bigger than one, run the regression imposing the restriction that the coefficients sum to one. The simplest way to do that is to follow the original regression with

Code: Select all

RESTRICT(CREATE) 1
# 1 2
# 1.0 1.0 1.0

With many restrictions, it's not as simple to do a check of the boundary. You would need use LQPROG to impose the constraints. There's an example of that in section 12.5 of the User's Guide.