DCC implementation. Not Reach an Optimal Solution

Dear All,

I am in attempt to work out on DCC GARCH implementation for 2 assets. My data consists of 2 stock return which are on weekly basis and I got a total of 670 observations. I apply the dcc garch toolbox from MFE Kevin Sheppard. I have tried this by using the following code where ;

[PARAMETERS,LL,HT,VCV,SCORES,DIAGNOSTICS]=dcc(data,[],1,0,1);

the program seems to work well but toward the end of the optimisation procedure, I got the following message :

Local minimum possible. Constraints satisfied.

fmincon stopped because the size of the current step is less than

the default value of the step size tolerance and constraints are

satisfied to within the selected value of the constraint tolerance.

Any recommendation for this issue and how can I tackle this problem.

[Walter Roberson]

This question appears to relate to the MFE Toolbox https://www.kevinsheppard.com/MFE_Toolbox

Answer 1

[Juan Esteban De la Calle Echeverry]

You indeed have the correct and optimal solution!

[Walter Roberson]

That is, fmincon is a local minimizer, and never promises to be a global minimizer. It has found a point that looks good as far as it can tell, but since it has not searched all possibilities, it cannot prove that you have found the best possible solution.

In some situations, the mathematics is such that you can show that there is only one minima. I do not know much at all about DCC models, but when I look at the information I find, it seems unlikely to me that that situation applies: I suspect that there are multiple minima.