Closed andreufont closed 2 weeks ago
@jchavesmontero - this is related to the fact that we do not pass something like X_model_fid to the extra theory in the likelihood.py module, like we do for the other nuisances (we pass F_model_fid=self.theory.F_model_fid
)
I guess following the other PR (#66 ) we also need a similar different duplication of models for the metals?
This is becoming relevant because I would like to add yet another model for HCD contamination, so we should probably sit down and discuss exactly how these models should be dealt with systematically.
In the branch fit_desi_edr
I solve this issue passing self.metal_models
to the extra theory, instead of passing include_metals
with labels. As I mentioned above, however, whenever we are both back to work we should discuss this.
Maybe we don't need to setup an alternative theory object for the high-resolution spectra. This was a hack that probably was a bad idea. It was motivated by the fact that the theory needed to know the redshifts from the data for efficiency, and high-res data could have different zs than low-res data. However, we could easily solve this by add a function called theory.update_zs
or similar that merges the redshifts needed by both datasets.
I removed the theory object for high-res data in my last commit. I had to specify zs here and there so it works.
The way we are dealing with the extra theory for high-res P1D is getting messy, I should take a look at this.