Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Add parameters correlation matrix property #1918
This PR moves the two-line method to compute the correlation matrix from the covariance matrix to the Parameters class, which currently holds the covariance matrix.
@adonath and I discussed this morning how to continue with the
The biggest question at the moment is where the parameter error information is stored and how it is exposed. Sherpa has it separately from the model here:
There's also many API questions how to expose error information on the
Is this PR OK or already controversial?
Having the covariance and correlation matrices living on
Parameters seems the most natural thing to me. Having it on a
Model would prove difficult when fitting inhomogeneous datasets simultaneously because they won't share exactly the same model.