Skip to content

Fix a missing factor of 2 in the Fisher Matrix proposals

Gregory Ashton requested to merge fix-factor-2 into master

Reviewing the bilby_mcmc Fisher Matrix Proposals, I realised we are missing a factor of 2 in using the inverse Fisher matrix as a covariance matrix. To test out the change, I reran the GW150914 review analysis with and without the change:

Analysis ACT Efficiency
Original analysis without factor of 2 210+/-90 0.065%
Analysis with factor 2 fixed) 180 +/- 80 0.072%

Hence, one sees a modest improvement in the efficiency. Meanwhile, all comparison plots still pass the JS test.

Merge request reports