davidthomas5412 / panglossnotebooks Goto Github PK

Reviewing the GenerateMockData1_2017_01_18 notebook I have some suggestions:

• We need to use the MS halo masses directly (and not resample them), in order to get the right masses in the right places

• Related to this, the Mhalo node in the mock data generation PGM should be a dot (for "determined by the Millennium Simulation)

• Nice work putting in a placeholder L-M relation! With a small modification it can be close to realistic, I think.

The numbers for the benchmark test in https://github.com/davidthomas5412/PanglossNotebooks/blob/master/GroupMeeting12_5_16.ipynb are encouraging (factor of 200 speed-up) but you should show that the two codes get the same answers too. Maybe you can avoid computing correlation functions etc by computing the same set of predicted background galaxy shears with the two codes, and investigating the component-wise differences? They should be at least very tightly correlated if not actually equal. Thoughts?