Hi, any help with this would be great:

I am optimizing a model containing different parameters and I want to use the Hessian matrix to find the errors of my parameters.

I know that:

a) Near the global minimum, the inverse of the Hessian matrix provides a good approximation of the covariance matrix for the independent model parameters.

b) The inverse eigenvalues of the Hessian are the errors of combinations of the estimated model parameters.

My questions are:

1. Does this make sense?

2. What are the units of the errors? What do these errors mean?

3. How can I calculate these errors? When I use [v,lam]=eig(hessian); in Matlab, the eigenvalues are ordered from smallest to largest, but I need to know which eigenvalue corresponds to which parameter.