search

scikit-hep / hepstats

Statistics tools and utilities.
https://scikit-hep.org/hepstats/
69 stars 13 forks source link

Bayesian blocking gamma prior doesn't make sense #83

Closed Moelf closed 1 year ago

Moelf commented 1 year ago

when gamma is supplied, https://github.com/scikit-hep/hepstats/blob/4c5bdba397b1241d4dbc1fa75ce76bdd6d9a9287/src/hepstats/modeling/bayesian_blocks.py#L51 https://github.com/scikit-hep/hepstats/blob/4c5bdba397b1241d4dbc1fa75ce76bdd6d9a9287/src/hepstats/modeling/bayesian_blocks.py#L154

this is a constant prior, this doesn't make sense because the prior should depend on N_b (https://arxiv.org/pdf/1708.00810.pdf): image