search

pierrexyz / pybird

Python code for Biased tracers in redshift space
https://pybird.readthedocs.io/en/latest/
MIT License
26 stars 19 forks source link

Bug for multiple skycuts in dev branch? #2

Closed cullanhowlett closed 3 years ago

cullanhowlett commented 3 years ago

Hi Pierre, Guido and all,

We've been looking at fitting multiple redshifts/skycuts using pybird and think we've identified a potentially important bug in the montepython class file, although it's not clear which branches this may be in, or if we've missed/modified something in our fork.

In correlator.compute, the redshifts are reordered such that the model is computed only for the middle bin and then rescaled by the growth factor appropriately for all other redshift bins. However, it looks the linear power spectrum from class is computed for the first redshift in the unordered list of redshifts specified in self.config (zfid in likelihood_class.__set_cosmo. If the middle redshift is different from the first redshift, this causes a mismatch between the linear power spectrum and growth factors.

pierrexyz commented 3 years ago

Hi Cullan, Yes it is possible that this is the dev branch that has seen multiple commits... I've checked that on the current commit the linear power spectrum is correctly computed for the middle redshift, see l. 847 in pybird_dev/pybird.py. We'll make it clean when merging with master, meanwhile I am sorry I know the dev branch is a bit messy with frequent commits...