search

jfkcooper / HOGBEN

Experimental design in neutron reflectometry using the Fisher information.
BSD 3-Clause "New" or "Revised" License
6 stars 3 forks source link

Are we double counting covariance? #18

Closed jfkcooper closed 2 years ago

jfkcooper commented 2 years ago

FI matrix gets summed for information, but it is symmetric so $g{i,j} = g{j,i}$. Check if we want to add them all up, or just take one half of the matrix and then implement.

jfkcooper commented 2 years ago

After discussion with Lucas:

I confused FI matrix with covariance matrix, FIM is not symmetric so initial assumption is wrong. Covariance matrix is symmetric, but is currently only used in data analysis. We use np.linalg.eigvalsh which seems to be fine as we are consistent. We could half the covariances if wanted, as in some ways there is a factor of two in there, but this would just result in optimizing a slightly different condition.