Add support for Python 3.11, drop support for Python 3.7
fix toy generation with constraints
Upgrade to Python 3.10 and zfit >= 0.10.0
Enhanced speed toy limit calculation
Add multidimensionl PDF support
Add support for binned data and models
Upgrade to Python 3.9 and drop support for 3.6
loss: upgrade API to use
create_newto make sure that the losses are comparable. Compatible with zfit 0.6.4+
sPlot: Increase the tolerance of the sanity check from 1e-3 to 5e-2, if above the tolerance a ModelNotFittedToData exception is raised. In addition if the the check is above the 5e-3 tolerance a warning message is printed.
New documentation style
hepstats can now do hypothesis tests, and compute upper limits and confidence intervals for counting analysis
Progess bars are used to see the progression of the generation of the toys
ConfidenceInterval can compute Feldman and Cousin intervals with boundaries (i.e
AsymptoticCalculator asymov weights are now scaled to the number of entries in dataset from loss function if the loss is not extended
hepstats.hypotests can now be used even if there is no nuisances. The pll function in utils/fit/diverse.py had to be modified such that if there are no nuisances, the pll function returns the value of the loss function.
add notebooks demos for FC intervals with the
add warnings when multiple roots are found in
move toys .yml files from notebook to notebook/toys
Redesigned packaging system, GHA deployment.
expected_poi removed from BaseCalculator and AsymptoticCalculator
add type checks in the hypotests submodule
hepstats is now compatible with zfit > 0.5 api
expected intervals in upper limit are now calculated from the pvalues and not from the expected_poi function anymore.
Addition of the sPlot algorithm
Addition of the FrequentistCalculator to performs hypothesis test, upper limit and interval calculations with toys. Toys can be saved and loaded in / from yaml files using the methods:
New version for the new hepstats name of the package
Package name changed from scikit*stats to hepstats
Additions of classes to compute upper limits and confidence intervals.
Release for Zenodo DOI
First release of scikit*stats
Addition of the modeling submodule with the
Bayesian Blocks algorithm
Addition of the hypotests submodule