12
$\begingroup$

I was wondering if those who used the entropy pooling code provided by Attilio Meucci had issues with the optimization procedure (especially regarding the fminunc function in Matlab). When I stress for example the expectation and the volatility, I get weird results for some stress values Stock price distribution with stress values 100 for expectation and 20 for volatility, prior in blue, posterior in red

and the following warning:

Local minimum possible:
fminunc stopped because the final change in function value relative to 
its initial value is less than the selected value of the function tolerance.

How can one deal with this ?

For information, the image represents a stock price distribution with stress values 100 for expectation and 20 for volatility, prior in blue, posterior in red.

$\endgroup$
  • $\begingroup$ Can you show us your data? $\endgroup$ – Bob Jansen Feb 28 '13 at 18:35
  • 2
    $\begingroup$ I have noticed that taking extreme views can tend to cause problems (the effective number of simulations being low is usually a good sign). One simple way to check the results is to assume it is a normal distribution and take the views using the analytical formula and see how the results change. $\endgroup$ – John Mar 1 '13 at 0:03
  • 2
    $\begingroup$ Thanks John, I compared the results with the normal distribution, and the numerical solution indeed does not match the analytical solution when I apply extreme views on expectation or volatility. For those interested, the solution is in section 4 of the "Fully flexible extreme views" paper, a Gauss-Hermite grid solves the problem. $\endgroup$ – Bytesize Mar 1 '13 at 14:26

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.