jaxgsaFind which inputs matter

Your first analysis ​

Suppose a model has three uncertain inputs. You want to know which ones explain changes in its output. The getting started guide takes you from defining those inputs to interpreting a Sobol result. In that result, S1 measures an input's contribution on its own; ST also includes its interactions with other inputs. A large gap between them tells you that the input matters in combination with others.

After the first run, Concepts explains the ideas behind those numbers. The basic example shows how to check whether the estimates have settled as you add samples.

Please cite ​

If you use jaxgsa in research, please cite it. For the software citation, use the general DOI 10.5281/zenodo.22099117, which covers all versions:

@software{pessina2026jaxgsa,
  author    = {Pessina, Daniele and Papathanasiou, Maria M.},
  title     = {jaxgsa: Global Sensitivity Analysis in JAX},
  year      = {2026},
  doi       = {10.5281/zenodo.22099117},
  publisher = {Zenodo},
  url       = {https://doi.org/10.5281/zenodo.22099117}
}

As a text citation:

Daniele Pessina and Maria M. Papathanasiou (2026). jaxgsa: Global Sensitivity Analysis in JAX. Zenodo. https://doi.org/10.5281/zenodo.22099117

To cite a specific release, use its version DOI instead — for v1.0.2 that is 10.5281/zenodo.23057233.

Where to go next ​

• Choose a method when you know the question but are unsure which analysis to run.
• Use non-uniform inputs when a range alone does not describe how likely each input value is.
• Plan a larger study when model evaluations are expensive or you have many inputs.
• Look up the API for arguments, return values, and supported input types.

Performance ​

Performance depends on the method, sample count, number of inputs, and output shape. See the benchmarks guide for measured workloads and hardware.

jaxgsa's Sobol sampling and analysis workflow follows SALib, reimplemented for JAX.