Partial derivatives of model flux with respect to a stellar parameter.

[gully]

Partial derivatives

The problem: A user might want to inspect, say, the gravity-sensitivity of a spectral region of interest by plotting the partial derivative of the model flux with respect to log g. Right now I'm accomplishing this by running a model, saving the output spec.json file, and then differencing it with a revised spec.json file.

Suggested solution: Some of this differencing could be put into a script, and optionally saved inside the spec.json, or some other file. splot.py could be enhanced to include plotting functionality.

[iancze]

If all you are after is intuition, and not some smooth function to be used in an optimization routine, I think this might be accomplished more easily by simply taking approximations to the derivative of the flux in each dimension of the grid.

This could be done via a new but similar class to HDF5Creator: iterate through the raw, hi res grid and calculate derivatives across each dimension and then store this in a new HDF5. Then perhaps create a three-panel plot for each parameter combination, or some other sophisticated way of displaying the derivatives.