Previous: What do I do

Up: Frequently Asked Questions

Next: What do I need

What do I do if the Hessian is not positive definite?

If you have covariates, make sure you didn't include a constant term or variables that are very highly collinear. Whether you included covariates or not, what often helps is to narrow the variance of the prior on $\sigma_b$, $\sigma_w$, and $\rho$ by setting _Esigma and _Erho to smaller numbers than their defaults. Following this action is also sometimes helpful in getting the model to fit appropriately, if ${\mathfrak{E}I}$ took several tries to compute a valid variance matrix. You should be careful, of course, that a non-positive definite Hessian doesn't indicate an impossible estimation problem. Also be careful whenever eiread's ghactual is other than {0 1} (if it is, see the global _EI_vc).