multiHGtest
Testing for differences in survival using multiple hypergeometric P-values
Overview
The multiHGtest package provides nonparametric methods for comparing survival between two populations with sensitivity to rare and localized departures from the null hypothesis of equal hazards. Unlike standard tests (e.g., log-rank), these methods are designed to detect non-proportional hazard differences that occur at only a few time points.
The approach works by:
- Computing exact hypergeometric P-values at each time interval, testing whether the observed number of events in one group deviates from what is expected under equal hazard rates.
- Combining these per-interval P-values using higher criticism or Fisher's combination test to obtain an overall test statistic.
This package implements the methods described in:
Kipnis, A., Galili, B., and Yakhini, Z. (2026). "Higher criticism for rare and weak non-proportional hazard deviations in survival analysis." Biometrika, 113(1).
Installation
From PyPI:
pip install multiHGtest
From source:
git clone https://github.com/alonkipnis/multiHGtest_package.git
cd multiHGtest_package
pip install -e .
Requirements
- Python >= 3.8
- numpy >= 1.20
- scipy >= 1.7
- pandas >= 1.3
- multiple-hypothesis-testing
Quick Start
From survival table data (at-risk counts and events)
import numpy as np
from multiHGtest import hchg_test, fisher_hg_test, hg_test_dashboard
# Survival data: at-risk counts and events at each time interval
Nt1 = np.array([100, 95, 90, 85]) # Group 1 at-risk counts
Nt2 = np.array([100, 92, 88, 82]) # Group 2 at-risk counts
Ot1 = np.array([5, 5, 5, 3]) # Group 1 events
Ot2 = np.array([8, 4, 6, 3]) # Group 2 events
# Higher Criticism test
hc_stat = hchg_test(Nt1, Nt2, Ot1, Ot2, alternative='two-sided')
print(f"HC statistic: {hc_stat:.4f}")
# Fisher combination test
fisher_stat, fisher_pval = fisher_hg_test(Nt1, Nt2, Ot1, Ot2)
print(f"Fisher statistic: {fisher_stat:.4f}, p-value: {fisher_pval:.4f}")
# Full dashboard with all statistics
df, stats = hg_test_dashboard(Nt1, Nt2, Ot1, Ot2)
print(df)
print(stats)
From time-to-event data
If your data is in the standard time-to-event format (e.g., from clinical trials), you can convert it first:
import pandas as pd
from multiHGtest import from_time_to_event_to_survival_table, hchg_test
# Time-to-event data with columns: time, event (1=event, 0=censored), group
data = pd.DataFrame({
'time': [1, 2, 3, 4, 5, 1, 2, 3, 4, 5],
'event': [1, 1, 0, 1, 0, 1, 0, 1, 1, 0],
'group': [1, 1, 1, 1, 1, 2, 2, 2, 2, 2]
})
# Convert to survival table format
result = from_time_to_event_to_survival_table(data)
# Run the HCHG test
hc = hchg_test(
result['at_risk_1'], result['at_risk_2'],
result['observed_1'], result['observed_2'],
alternative='two-sided'
)
print(f"HC statistic: {hc:.4f}")
API Reference
hchg_test(Nt1, Nt2, Ot1, Ot2, alternative='two-sided', **kwargs)
Higher Criticism test of hypergeometric P-values. Returns the HC test statistic.
| Parameter | Description |
|---|---|
Nt1, Nt2 |
At-risk counts per time interval for groups 1 and 2 |
Ot1, Ot2 |
Event counts per time interval for groups 1 and 2 |
alternative |
'greater', 'less', or 'two-sided' (default) |
gamma |
HC threshold parameter (default: 0.4) |
randomize |
Use randomized P-values (default: False) |
fisher_hg_test(Nt1, Nt2, Ot1, Ot2, **kwargs)
Fisher combination test of hypergeometric P-values. Returns (statistic, p_value).
hg_test_dashboard(Nt1, Nt2, Ot1, Ot2, **kwargs)
Comprehensive dashboard returning a DataFrame with per-interval results and a dictionary with HC, Fisher, and MinP statistics.
hypergeom_test(k, M, n, N, alternative='greater', randomize=False)
Exact hypergeometric test for comparing proportions. Supports 'greater', 'less', and 'two-sided' alternatives.
from_time_to_event_to_survival_table(data, time_col, event_col, group_col)
Convert time-to-event (survival) data into the at-risk/event count format required by the test functions.
Legacy aliases
For backward compatibility: HCHGtest, FisherHGtest, testHG_dashboard.
Running Tests
pip install -e ".[dev]"
pytest
Citation
If you use this package in your research, please cite:
@article{kipnis2026higher,
title={Higher criticism for rare and weak non-proportional hazard deviations in survival analysis},
author={Kipnis, Alon and Galili, Ben and Yakhini, Zohar},
journal={Biometrika},
volume={113},
number={1},
year={2026}
}
License
MIT License. See LICENSE for details.
Metadata
Release files for multiHGtest 0.0.6
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| multihgtest-0.0.6.tar.gz | 15.5 kB | Without provenance | Details for multihgtest-0.0.6.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| multihgtest-0.0.6-py3-none-any.whl | Python 3 | none | any | Details for multihgtest-0.0.6-py3-none-any.whl |
Total release size: 23.9 kB
Release files / multihgtest-0.0.6.tar.gz
| Download URL | multihgtest-0.0.6.tar.gz |
|---|---|
| Size | 15.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
e49d25118bc90ba8778b642fee2202652943a2a8d2a802e9f81083e42cfdc0f4
|
|
BLAKE2b-256 checksum How to use checksums |
90482da735b587b0641d105997d76a67e52a491c1c653ef241b382a45a38cedd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.6
|
Release files / multihgtest-0.0.6-py3-none-any.whl
| Download URL | multihgtest-0.0.6-py3-none-any.whl |
|---|---|
| Size | 8.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
2cffb0c14a36e22f397b5ac2346e3c1a75bba560428335232fbaeab5bc541742
|
|
BLAKE2b-256 checksum How to use checksums |
7b4f3fc3142b7249bc864b658bc353c7f08cb155260412a7051ef3dc1baaf18a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.6
|
Security for multiHGtest 0.0.6
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.