sensitivity analysis

517 views
Skip to first unread message

DucMinh Nguye

unread,
Jan 24, 2014, 2:37:58 PM1/24/14
to ai...@googlegroups.com
Dear,

After running model successfully, I'd like to execute the post-analysis. How can I have a look the Table of sensitivity analysis in AIMMS that show the information about shadow price and reduced cost  

Thanks

Best regards,
DucMinh

Guido Diepen

unread,
Jan 26, 2014, 12:55:59 PM1/26/14
to ai...@googlegroups.com
Hi DucMinh,

Search for sensitivity analysis through this group resulted in the article https://groups.google.com/d/topic/aimms/tHZOKKkREas/discussion 

Guido Diepen

DucMinh Nguye

unread,
Jan 28, 2014, 10:20:04 AM1/28/14
to ai...@googlegroups.com
Hi,

I wonder that according to the definition of "shadow price", it is the marginal value of change of objective function from one unit increase in its right-hand side. The nonbinding constraint has its shadow price equal to zero, meaning even you increase the value on the right hand side of the nonbinding constrains, the optimal value is unchanged. However, in my model the shadow price of one constraint is zero but if I increase the value of parameter on the right hand side of this constraint, the optimal value also changed. That makes me confused and do not know how to conclude. 

Can you share some experience about this case. In my model, there is only binary variables.

Many thanks.

Best regards,
DucMinh Nguyen


On Friday, January 24, 2014 8:37:58 PM UTC+1, DucMinh Nguye wrote:

Peter Nieuwesteeg

unread,
Jan 28, 2014, 8:11:01 PM1/28/14
to ai...@googlegroups.com
That is the common definition of the shadow price. However, it assumes that the basic status of the variables and constraint doesn't change. So, in some cases it might just be valid for a small change in the right hand side of a constraint and not for one unit of change.

In your particular case it is related to the fact that you have a MIP model. The shadow prices of MIP models are calculated during the post solve. In the post solve, all integer variables are fixed and the shadow prices are calculated the resulting LP. As you only have binary variables, this LP is actually empty and hence the shadowprice will be 0. Your objective still changes when you change the right hand side, because your binary solution actually changes, but that is not covered by the shadow price.

Peter Nieuwesteeg
Senior AIMMS & Optimization Specialist
Reply all
Reply to author
Forward
0 new messages