Bayesian Blocking prior is astro-specific

scikit-hep / hepstats

Statistics tools and utilities.

69 stars 13 forks source link

Bayesian Blocking prior is astro-specific #80

Closed Moelf closed 1 year ago

Moelf commented 1 year ago

https://github.com/scikit-hep/hepstats/blob/85cdb1e9c0a2699fbe4715ff7c879f068ca7452f/src/hepstats/modeling/bayesian_blocks.py#L53-L54

Moelf commented 1 year ago

https://arxiv.org/pdf/1304.2818.pdf should probably use this, the 2012 has wrong formula

eduardo-rodrigues commented 1 year ago

Thanks @Moelf for the comment! You are most welcome to submit a PR otherwise we will get to this asap.

Moelf commented 1 year ago

no sorry I think you guys already used the correct formula, but I think now the problem is we shouldn't default to the prior one anyway, we should use the geometric prior, because the formula in screen shot comes from fitting astrophysics data, our histograms have nothing to do with brightness sequence of stars

jonas-eschle commented 1 year ago

Hi @Moelf , I am myself not an expert in this, however it's not quite clear to me why a geometric prior should be used? But as @eduardo-rodrigues mentioned, you're very welcome to open a PR and we can discuss it there (I have too little time currently to dig deep enough into the topic).

Moelf commented 1 year ago

Because the prior and numbers you're currently using is derived from fitting a simulation of star light, nothing to do with HEP