tausurv

Survival analysis for Python: nonparametric estimators, regression models, tree ensembles and neural networks behind one interface, with support for competing risks.

Quickstart Tutorials Concepts API reference

A Kaplan-Meier curve with a 95% confidence band and an at-risk table, on the Mayo Clinic PBC trial:

import tausurv as ts

ts.plot.set_style("publication")
X, Y, delta = ts.datasets.load_pbc()
ts.plot.km(Y / 365.25, delta, xlabel="years from registration")

Kaplan-Meier survival curve on the PBC trial cohort, with a 95% confidence band and an at-risk table below the time axis.

Contents of the package

subpackage	contents
`ts.nonparametric`	Kaplan-Meier, Nelson-Aalen, Aalen-Johansen, censoring distribution
`ts.linear`	Cox proportional hazards, Weibull, log-normal and log-logistic AFT, Fine-Gray
`ts.trees`	survival tree, random survival forest, gradient boosting
`ts.nn`	DeepSurv, DeepHit, logistic hazard, deep survival machines, CopulaSurv, HACSurv
`ts.metrics`	concordance, time-dependent AUC, Brier score, calibration, with cause-specific variants
`ts.model_selection`	cross-validation, hyperparameter search, nested cross-validation
`ts.datasets`	100 survival tables from 68 studies, downloaded on first use and checksummed
`ts.simulations`	single-risk and competing-risks data with known truth
`ts.plot`	survival and incidence curves, forest plots, calibration, metrics over time, SHAP

Every fitted model exposes the same prediction methods: survival function, cumulative hazard, cumulative incidence, restricted mean survival time and a risk score.

Documentation

Tutorials work through an analysis of the PBC trial: Kaplan-Meier curves, a Cox model with held-out evaluation, and competing risks.
How-to guides solve one task each.
Concepts explain the quantities and assumptions behind the methods.
Plotting covers the plot functions and styles.
API reference is generated from the docstrings.