Metrics for physicsmodel grid quality

[karllark]

The quality of our sampling of parameter space is not well defined. We visually decided that the PHAT production parameters were "good enough." The production grid was based on isochrones as the base, hence are non-uniform in mass, but uniform in age. One idea to make better sampled grids is to go directly from the evolutionary tracks resulting in non-uniform age and uniform mass.

How to decide which sampling is best?

One idea that was discussed at the 9 Nov 2018 HackDay (discussions between @karllark and @lcjohnso) would be to use the density of points on a HR diagram (Teff versus Luminosity) as a criteria, going for uniformity. This would need to be checked in observed space (e.g., CMDs) as well given things may change fast in one observed band, but not others. Quantifying the density could be done by tabulating the min/max/median change in parameters/observables in for different grid creation methods.

[karllark]

Related to #18.

[karllark]

Current thoughts/work specific to coverage of stellar parameters. May wish to think about this for dust parameters as well (currently uniform, but may be better to have some other spacing.

[benw1]

May be worth making the grid spacing take the observed fluxes into account. There is already a push to avoid putting in multiple models that sample the same fluxes. One could also imaging sampling models more densely at fluxes that have higher signal-to-noise.