[ADMB Users] how reliable are ADMB estimates of variances?

[Richard.Methot at noaa.gov]

Generally we find ADMB's estimates of variance from the inverse 
Hessian method to be sufficiently accurate.  Even with large and 
complex models with hundreds of estimated parameters, we find these 
variance estimates to be similar to those calculated from MCMC and 
from likelihood profiles.

Rick Methot

----- Original Message -----
From: Saang-Yoon Hyun <HYUS at critfc.org>
Date: Sunday, July 12, 2009 2:34 pm
Subject: [ADMB Users] how reliable are ADMB estimates of variances?

> Hello.
> 
> A simple ADMB program of mine does an excellent job of calculating 
> MLEs of parameters.  However, its estimation of variances of 
> parameter estimates does not look good.  If someone has ever 
> evaluated ADMB performance in terms of "variance estimation", 
> please share your findings/experiences/opinions about the 
> performance with me.  I wonder about how reliably ADMB calculates 
> variance estimates.
> 
> Thank you, 
> 
> Saang-Yoon
> 
> _______________________________________________
> Users mailing list
> Users at admb-project.org
> http://lists.admb-project.org/mailman/listinfo/users
>