[ADMB Users] eigenvectors of the Hessian and parameter errors

[David Holmgren]

Dear ADMB list,

I am using ADMB 9.0.x on a Windows XP machine, and I have been working on a problem in which I expect to see an indefinite Hessian.  In other words, when I run the compiled program, I see two warnings about an indefinite Hessian at the end of the output.  I understand that in such cases parameter errors are not computed.  However, if I specify -eigvec on the command line, I find that the parameter errors are in fact computed!  Is this indeed correct behaviour, and why is it occurring?  My apologies in advance if this question sounds a bit odd, but I am just getting up to speed with ADMB.

Many thanks.

Sincerely,

David Holmgren

David Holmgren
holmgren at telus.net