nsmith- / rhalphalib

A binned fit intermediate representation library

rhalphalib

Quickstart

First, install Combine v9 using your choice of installation instructions (with CMSSW, using LCG, or inside a Conda environment), then in that environment, run:

python3 -m pip install --user https://github.com/nsmith-/rhalphalib/archive/master.zip

Take a look at test_rhalphalib.py for examples of how to use the package.

rhalphalib's Issues

Defer systematic morphing to combine for ParametricSample

Now that cms-analysis/HiggsAnalysis-CombinedLimit#634 is available we can let RooParametericHist take care of the morphing rather than building it ourself.

1xN vs Nx1 array handling in BernsteinPoly

 rl.BernsteinPoly("tf", (2, 0), ['rho', 'pt'],  limits=(0, 10))

is understood correctly, but

 rl.BernsteinPoly("tf", (0, 2), ['rho', 'pt'],  limits=(0, 10))

throws an error. It seems the fcn mishandles an array like [[1], [1], [1]] though a 1D array like [1,1,1] or a proper 2D array is fine.

Smoothstep implementation is wrong

From Thiago, the formula at

rhalphalib/rhalphalib/parameter.py

Line 280 in cdab8c8

formula = "(((0.1875*@0*@0 - 0.625)*@0*@0 + 0.9375)*@0 + 0.5)*TMath::Sign(1, 1+@0)*TMath::Sign(1, 1-@0) + 1 - TMath::Sign(1, 1-@0)"

doesn't behave as it should, blowing up at large values.
I think (((0.1875*x*x - 0.625)*x*x + 0.9375)*x + 0.5)*(x > -1)*(x < 1) + 1*(x >= 1) is what it should be.

[Feature Request] Incorporate MC Statistical Uncertainties

• Extend TemplateSample class to allow automatic incorporation of MC statistical uncertainties
• The simplest way would be: add a bin-by-bin error member which is used to construct nbins up and down templates with independent nuisance parameter names (to allow uncorrelated up/down fluctuations)
• A better way would be to do some kind of Barlow-Beeston lite approach (which is not currently supported in combine with RooDataHist templates, but works if you use histograms)

Install with CMSSW+combine is missing flake8

Installing rhalphalib according to the README with CMSSW+combine, I get this error: pkg_resources.DistributionNotFound: The ‘flake8’ distribution was not found and is required by the application

If I run this before installing rhalphalib, it is fine:
pip install --user flake8