How does fmincon works? Changing not binding restrictions changes the result

Hello.

I am working with the function fmincon and I have noticed that changing a restriction can alter the result even when this restriction is not binding.

For example, I am using the code

[theta] = fmincon('gmmsimobj2',theta0,[0 -5 -1],[0],[],[],[0.001 -1 0],[1 0 1],[],foptions,simfunc,e,sp,beta,momfunc,y,mp,alg,W);

Which gives me a result of theta=[0.0177 0 6.8070e-005].

Then I change some restrictions that were not binding before

[theta] = fmincon('gmmsimobj2',theta0,[0 -5 -1],[0],[],[],[0.001 -inf 0],[inf 0 inf],[],foptions,simfunc,e,sp,beta,momfunc,y,mp,alg,W);

This gives a result of theta=[0.0148 -7.7908e-004 0.0039], which in fact was a possible answer in the last excercise (it did not violate any of the previous restrictions).

Could you tell me why the result changes? I do not know exactly how this function works. How can these changes in the restrictions affect the result?

Thank you! Javier

[Matt J]

You should ALWAYS call fmincon with at least 4 output arguments

 [theta,fval,exitflag,output] = fmincon(...)

When you do, what are the values of the out arguments in each case?