0
$\begingroup$

I am currently solving a Non Linear modell in GAMS and I am interested in a sensitivity analysis of the results. When working with a linear program I am able to look at the marginal values for the constraints (a marginal value of 100 means that if the constraint is relaxed by one unit then my objective will increase with 100 units).

However when dealing with non linearity I have been informed that this is not the case and I am interested in finding a similar method of analysis for my non linear modell. How should I think?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

The marginal values you're describing are duality multipliers. You should be able to obtain duality multipliers from SNOPT in GAMS output; they should still be called marginal values, and associated with each constraint equation.

Duality theory for nonlinear programming is different than for linear programming. In linear programming, strong duality always holds, whereas in nonlinear programming, strong duality requires technical conditions that are not always satisfied (typically, you need a convex problem plus a constraint qualification, like Slater's condition, or LICQ). Weak duality always holds, though. (I'm also assuming your problem is smooth; if it's not, the situation becomes even more complicated.)

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.