# jaxgsa

> Global sensitivity analysis in JAX

jaxgsa tells you which inputs drive variation in a model's output. You choose
input distributions, evaluate your model at sample points, and use jaxgsa to
estimate each input's effect. It does not run your model for you.

Analyses are JIT-compiled in JAX and run on CPU, GPU, or TPU. Fourteen methods
cover variance-based indices (Sobol, eFAST, Kucherenko), screening (Morris,
DGSM), surrogate-based analysis (PCE, RS-HDMR, Shapley effects, VKOGA), and
given-data measures (HSIC, PAWN, Borgonovo delta, optimal transport, rank-based
first-order indices).

## Documentation

- [Home](https://danielepessina.github.io/jaxgsa/): Project overview, quick start, and citation.
- [Getting started](https://danielepessina.github.io/jaxgsa/guide/getting-started.html): First Sobol analysis, step by step.
- [Choosing a method](https://danielepessina.github.io/jaxgsa/guide/methods.html): Which of the fourteen methods fits your question and data.
- [Concepts](https://danielepessina.github.io/jaxgsa/guide/concepts.html): What sensitivity indices mean and how to interpret them.
- [API reference](https://danielepessina.github.io/jaxgsa/api/): Full exported surface, parameter contracts, and result fields.
- [Examples](https://danielepessina.github.io/jaxgsa/examples/basic.html): Worked examples for every method and workflow.
- [Benchmarks](https://danielepessina.github.io/jaxgsa/guide/benchmarks.html): Measured performance for representative workloads.

## Source

- [GitHub repository](https://github.com/DanielePessina/jaxgsa): Source code, issue tracker, changelog, contributing guide.
- [PyPI](https://pypi.org/project/jaxgsa/): pip and uv installation.
- [Zenodo DOI](https://doi.org/10.5281/zenodo.22099117): Software citation for research use.

## jaxgsa studio

- [jaxgsa studio](https://danielepessina.github.io/jaxgsa/studio/): Run jaxgsa in the browser without installing Python.

## Agent skill

- [Skill: SKILL.md](https://github.com/DanielePessina/jaxgsa/blob/master/skills/jaxgsa/SKILL.md): Guidance for writing or reviewing code that uses jaxgsa.
- [Skill reference: variance-based](https://github.com/DanielePessina/jaxgsa/blob/master/skills/jaxgsa/reference/variance-based.md)
- [Skill reference: moment-independent](https://github.com/DanielePessina/jaxgsa/blob/master/skills/jaxgsa/reference/moment-independent.md)
- [Skill reference: screening](https://github.com/DanielePessina/jaxgsa/blob/master/skills/jaxgsa/reference/screening.md)
- [Skill reference: dependent-inputs](https://github.com/DanielePessina/jaxgsa/blob/master/skills/jaxgsa/reference/dependent-inputs.md)