search

LSSTDESC / txpipe-reanalysis

Re-analyze precursor surveys
3 stars 1 forks source link

Test firecrown with des y1 3x2 #9

Open chihway opened 4 years ago

chihway commented 4 years ago

List of things to check:

  1. make sure example runs and produce chain
  2. go through config file to understand what is what, also for output files
  3. make sure we can submit an actual batch job
  4. check single-likelihood evaluation
  5. check if can run multinest, of what does run-cosmosis mean
  6. free up more parameters to see what happens
  7. plot chains and output files
chihway commented 4 years ago

For the record, Firecrown is working in both Fisher and sampling mode now. And at the 2-parameter level they agree pretty well with cosmosis. When freeing all cosmo params it doesn't quite recover the truth, but this is still under investigation. This will mostly be tested in the cosmodc2 repo: https://github.com/LSSTDESC/txpipe-cosmodc2

image

image

katrinheitmann commented 4 years ago

I am really inside for too long ... when reading your comment "When freeing all cosmo params" I was thinking, how lucky they are that they are set free ...