I am working in Gauss 12 with Optmum module and I need to estimate the variance-covariance matrix. I wanted to take th inverse of the final Hessian calculated by Optmum, or just to take the inverse of _opfhess.
I tried and the results are too small, so I wonder that probably the entries of the Hessian matrix are multiplied by the number of observations?
Should I first to correct and divide my Hessian by the number of observations and only afterwards to take the inverse?
I tried to find the information in the manuals but I couldn't.
Thank you a lot in advance!
The Optmum examples, opt2.e and opt4.e show how to compute the covariance matrix of the parameters using the cross-product of the gradient matrix. You could also compute the Hessian using the hessp() function in the Run-Time Library. The final approximations from the iterations is not good enough for a covariance matrix. It's available primarily for diagnosing convergence problems.