Covariances

[DouglasBoubert]

In principle, any continuous numeric quantity might have a covariance with another continuous numeric quantity. The most pressing example is that Gaia will give the covariances between the five astrometric quantities. We should store correlations rather than covariances, because these are nice and dimensionless.

We could define a new quantity for each new possible correlation, i.e. create CORR_PROPERMOTIONRA_PROPERMOTIONDEC, however this could be tedious. Are there cleaner ways to do this, either in faststars or upstream in astrocats?

[guillochon]

I think what makes sense is to add a correlations attribute to the quantity, which is a dictionary containing a list of variables that variable is correlated with.

So for instance:

"propermotionra":{
   "value":"22",
   "correlations":{
      "propermotiondec":"0.3"
   }
}

That'll keep things clean and avoid needing to create a bunch of new keys. We can write this into the faststars module for now but it probably makes sense to migrate it up to astrocats in general in the future.

[DouglasBoubert]

I like that solution! We'll need to add checks to make sure if A is correlated with B then B is correlated with A. Could you add it to faststars and then I'll write the checks?