Alexander Held
Alexander Held
> the NPs for various contributions won't get combined into a single NP, resulting in unnecessarily slower fits If you use the same parameter names, the modifiers are going to...
Maybe to be practical, I think there are multiple aspects to this: - change the default from Gaussian to Poisson, - make the constraint term configurable, - include the constraint...
Related to roundtrips, an adversarial example with multiple `staterror` modifiers acting on the same sample & channel: #1830
See also #566 which includes an example. I have seen the `` tag being kept in these cases as well, without any path information provided. When considering the addition of...
related: #994
Related to `get_par_slice` above, `model.config.par_slice(par_name)` already exists.
related: https://github.com/scikit-hep/pyhf/discussions/1832 (`hypotest` behaves differently)
That is the option I thought of as well. Alternatively you could also of course keep the current approach where users just need to ensure they do not modify the...
Similarly to this, `suggested_init`, `suggested_bounds` and `suggested_fixed` currently take no arguments either. One other consideration is that maybe some of these parts of the API are meant to eventually support...
In this context it might also be nice to include the approximation via best-fit POI value divided by its uncertainty.