Hello everyone,
I want to make a sensitivity analysis of a given parameter. GAMS only gives the reduced costs of variables and shadow prices of constraints. I want to see what value of change of this scalar parameter can change the optimal basis? What should i do? Thank you in advance.
Turn the parameter into a fixed variable and you should get dual information. This might change the model type of your model. -Michael
Thanks for your reply.
My model is MIP, using a fixed variable will make it NLP as you mentioned. Can I do that in MIP or LP, without changing my model type?
Nope. If you do that (turn the variable again into a parameter via holdFixed and tryLinear) you do not get the dual for the variable/parameter. However, you can solve the resulting MINLP with DICOPT (with your favorite MIP solver). Since you have a problem with discrete variables you should be aware that the dual in GAMS ar the dual of the fixed problem (a MIP does not really have easy to compute duals). So you don’t get the impact of a variable change to the original problem but to the problem with the discrete variables fixed. Is this what you want?
-Michael
Thanks for your quick reply. Thats what I want.