At lunch today, Angie Wolfgang (PSU) gave a talk at the CCA on hierarchical inference of small and rocky exoplanet population properties. She made a nice set of arguments for the hierarchical Bayesian methodology, which was preaching to the converted (but good). She showed her results on exoplanet compositions and H/He envelopes, both of which are impressive, and then she went on to look at parametric and non-parametric fitting of the mass–radius relationship at small radii. She excludes zero mass scatter at fixed radius at all except the smallest radii. There is a consistent story emerging that the very smallest planets are indeed (pretty much) all rocky.
I had some back-and-forth with Megan Bedell (Chicago) about derivatives of our spectral model with respect to parameters. She has this all correct but I recommended parameterization changes, and the whole time, in the background, Dan Foreman-Mackey (UW) was saying things like “you should never take your own derivatives” and “you should use Theano”. I ignored him, probably at my peril.