to keep in mind for future development for general survey case - warning flags/integrity of labels

[mkness]

At some point we may want to assign some numbers/warning flags about the integrity of the labels - this could come from chi2 but with only 3 labels as an approximation is this a good option ? (4 labels might be enough though). Regions that are independent of the labels could = a higher chi2 if they deviate from reference objects (i.e regions of individual abundances not covariant with [Fe/H], [alpha/Fe]). For The Cannon paper I have excluded a small subset of stars in the Figures of stars on the isochrones, using APOGEE STAR BAD flags; these stars lie in non physical spaces. In general we should probably not rely on external warnings already existing?

[davidwhogg]

eventually we want to go Bayesian; in principle Bayes will make this all sensible and easy. Perhaps we should open an issue about going Bayesian.