Lorenz Gaertner

[WIP] feat: Personal applicataion of custom modifiers

1

# Description This is my personal use of a custom modifier. It includes some additional features: - usage of custom function - setting of custom parameter type - output of...

fix: Catch invalid model specification with duplicate channels, samples and modifiers.

1

# Description Fixes #2059 but extends to channels and samples. We want to catch the case where we have multiple channels with the same name in the model specification, where...

Treating arbitrarily correlated parameters with pyhf

3

### Summary So far, pyhf does not have the option to arbitrarily correlate modifier parameters. I have seen multiple workarounds implemented for Belle II analyses (and maybe also other places)....

Add a pyhf example (see #807). TODOs: - [ ] fix renaming through `parameter_map`. This is not actually renaming the parameter, but only the qualified name. - [ ] fix...

All `SignalPDF`s should have a physical observable as the numerator and their integral as the normalization

1

I noticed that a `B->pilnu::d^2Gamma/dq2/dcos(theta_l)` `SignalPDF` exists, but there is no corresponding `Observable`. It would be nice to have an `Observable` for every `SignalPDF`, the integral of which yields the...

Memory occupied by eos objects not freed after variables go out of scope

2

Hi, I noticed that eos occupies a lot of memory when successively creating analysis objects, even when I expect the memory to be freed after the variable goes out of...

Warn when parameter hits bounds

3

### Summary Hey, I recently ran a lot of toys for a very large model and had a hard time debugging, as some nuisance parameters hit the parameter bounds for...

fix: respect fixed parameters in n_dof calculation

1

Solves #478 by passing fixed parameters to `model_utils.unconstrained_parameter_count`.

Wrong `n_dof` for fixed unconstrained parameters in `_goodness_of_fit`

6

Hey, I think I found a small bug in the `_goodness_of_fit` function, when working with fixed unconstrained parameters. When you fix an unconstrained parameter in the fit using the `fix_pars`...