csminwel: Correct for low angle
Posted: Thu Oct 10, 2013 5:15 pmHello,
I have a question concerning the optimizer "csminwl" for the posterior mode computation: Does anybody know what the warning message "Correct for low angle" indicate? Does it indicate that the posterior mode is very flat at this point? (I have experienced that every time this warning message appears, the optimizer is very slow, i.e. the improvements for each iteration are very small. In this case, the mode is not found even after 1000 iterations.)
If this is the reason, one solution to this problem would be to tighten the priors (especially for parameters which seem to be very flat around the posterior mode)?
I am thankful for any help,
Best,
Niklas
I have a question concerning the optimizer "csminwl" for the posterior mode computation: Does anybody know what the warning message "Correct for low angle" indicate? Does it indicate that the posterior mode is very flat at this point? (I have experienced that every time this warning message appears, the optimizer is very slow, i.e. the improvements for each iteration are very small. In this case, the mode is not found even after 1000 iterations.)
If this is the reason, one solution to this problem would be to tighten the priors (especially for parameters which seem to be very flat around the posterior mode)?
I am thankful for any help,
Best,
Niklas