CLsb options

[cranmer]

Description

The Brazil band plot and the pyhf.infer.hypotest API is very tailored to CLs. It’s not clear how to get expected bands for CL_sb (standard frequentist limits) or CL_b (for expected significance / discovery). Would be nice to have an option for the standard CLsb limits and to make which is being used explicit in the API.

Probably having an option like use_CLs=True as a default would be sufficient. Alternatively, upper_limit_type="CLs" could work.

Is your feature request related to a problem? Please describe.

The astro community uses 90% CLsb upper limits mainly.

Relevant Issues and Pull Requests

This is also relevant for the test inversion / interval API #953

Additional context

I also am not a huge fan of the bad physics jargon of justing “CL” instead of p-value, but that’s a different issue.

[cranmer]

closing as this issue is also picked up in #966

[matthewfeickert]

I'll actually reopen this, so that we can track the PRs that close it.