search

Chris-Pedersen / LaCE_manager

Cosmological emulator for the 1D flux power spectrum of the Lyman-alpha forest
3 stars 1 forks source link

Bugfixes - 1 significant bug in the CMB prior, 2 more minor bugs in sampler and likelihood #43

Closed Chris-Pedersen closed 2 years ago

Chris-Pedersen commented 2 years ago

Identifed a few bugs to fix, starting with the most serious:

  1. When including the CMB likelihood, the neutrino mass likelihood parameter wasn't being passed to the cmb_like object, so changes to neutrino mass were not being considered in the prior evaluation. This is why we saw an abnormally flat CMB prior in neutrino mass when running the sampler on CMB only. This bug was consistent across all compression tests so shouldn't have any implications on our conclusions so far. I've set the compression chains to rerun with the fix, I think we can just expect that all the constraints tighten, we will just have a better estimate of the CMB prior to compare the P1D contribution to.
  2. Fixed a bug where you can set up a likelihood object with use_sim_cosmo=True, and it will use the mock simulation cosmology as the true cosmology, for testing purposes. This wasn't being propagated to the theory objects before.
  3. Fixed a minor bug in the sampler when running without parallelisation.