search

andycasey / smhr

Spectroscopy Made Hard(er)
14 stars 7 forks source link

MultiProfileFittingModel #178

Closed alexji closed 7 years ago

alexji commented 8 years ago

Because of the issues brought up by #152, I think it's a good idea to extend ProfileFittingModel to allow more than one component to be fit for blended lines (analogous to how quasar absorption people fit multiple velocity components in their absorbers). This allows fitting two overlapping Gaussians. The EW uncertainty derived can then include uncertainties in the parameters of the blending Gaussians.

The reason to add this (rather than making a synthesis spectral model) is that we don't have to know the identity of the blending line for this to work, while in synthesis you have to know what the blending line is.

andycasey commented 8 years ago

The current ProfileFittingModel actually does fit Gaussian profiles to neighbouring regions. It just doesn't place tight constraints on the joint parameters (e.g., one profile should have a similar \sigma as a neighbouring profile).

Depending on the goal, this can really be a time sink to get it generically right even for most circumstances. I think it should really be clear that this is actually the solution required for -- whatever the problem is.

alexji commented 7 years ago

Closing because we have decided this is not the best solution (better is to actually synthesize the blended line)