Open drphilmarshall opened 7 years ago
@davidthomas5412 The notebook that addresses this issue should contain a PGM describing what we are doing: let's use that diagram to be clear about what assumptions we are making (eg about the halo concentrations!). If you took a photo of the group meeting whiteboard, feel free to use that (you can make <img href="">
calls from markdown cells. Otherwise, issue me with making the simplest possible inference PGM and I'll provide some daft
code.
@davidthomas5412 Let's start with the simplest possible inference: simple Monte Carlo. We'll need the ForegroundCatalog to be able to
set_prior
and thendraw_halo_masses
, and then we'll need tocompute_likelihood
(using the the function @sweverett already implemented). This will give usN_samples * (N_halos + 1)
numbers (if we keep the concentrations fixed for now): this is going to get into the Mb quite quickly, so we might want to think about how to store the samples (we need to keep them all). A quick and easy solution is to write out simple plain text halo catalogs that can be read back in with the existingForegroundCatalog
method, but this won't scale well: let's start thinking about database table design.