Add function to convert to a LI prior

I understand what you were saying about the fraction of samples now. The issue is about sampling with/without replacement.

I would suggest that we do posterior.sample(frac=fraction, weights=weights, replace=True) and also compute the effective sample size, effective_sample_size = sum(weights)**2 / sum(weights**2).

We could add a warning if fraction > effective_sample_size / len(posterior).

I guess there are multiple things going on here, but when using dynesty we already resample with replacement, so this would just be another instance of that. At the least, I would suggest that replace should be an argument of this function.

added 1 commit

• 5bd5eadc - Add in replacement and a warning

Compare with previous version

@colm.talbot I've added your suggestions

approved this merge request

added 7 commits

• 5bd5eadc...3d257816 - 4 commits from branch master
• 738cfecd - Fix bug and start work on a test
• c745dcca - Merge branch 'master' of git.ligo.org:Monash/tupak into add-prior-conversion
• d85e2e6f - Merge branch 'add-prior-conversion' of git.ligo.org:Monash/tupak into add-prior-conversion

Compare with previous version

added 1 commit

• 8f56e8e6 - Adds a test

Compare with previous version

mentioned in merge request bilby_pipe!226 (merged)

approved this merge request

approved this merge request

approved this merge request

added Feature-request Priors labels

changed milestone to %0.6.0

merged

mentioned in commit 852ab635