Decision curve analysis for outcomes recorded with error. This is the Python
version of the R package in ../r; the tests check that both give the
same numbers.
pip install "git+https://github.com/Marwan-Alqumbaey/truedca#subdirectory=python"Requires Python 3.9 or later with NumPy, SciPy and pandas (requirements.txt);
plot_dca also needs Matplotlib.
Functions: nb_curve, dca_window, warp_threshold, nb_adjust, nb_bounds,
nb_bias_analysis, pilot_v, twophase_design, design_gain, twophase_dca,
make_strata, plot_dca. They take NumPy arrays and return pandas data frames.
Functions that use random numbers take rng (a seed or numpy.random.Generator).
In twophase_dca, truth is NaN for patients whose chart was not reviewed.
pip install -r requirements-test.txt
pip install .
pytestThe reference values in tests/fixtures come from the R package
(Rscript tests/fixtures/make_fixtures.R, run from this folder).