salib

Global sensitivity analysis for Rust, implemented from the primary literature. Bit-deterministic by construction.

Quickstart

[dependencies]
salib = "0.1"

use std::f64::consts::PI;
use salib::*;
use salib::samplers::{SobolSampler, build_saltelli_matrix};
use salib::estimators::estimate_saltelli2010;

fn main() {
    let problem = ProblemBuilder::new()
        .factor("x1", Distribution::Uniform { lo: -PI, hi: PI })
        .factor("x2", Distribution::Uniform { lo: -PI, hi: PI })
        .factor("x3", Distribution::Uniform { lo: -PI, hi: PI })
        .build()
        .unwrap();

    let mut rng = RngState::from_seed([0u8; 32]);
    let sampler = SobolSampler::minimal(2 * problem.dim());
    let saltelli = build_saltelli_matrix(&sampler, 8192, false, &mut rng).unwrap();

    // Ishigami function: y = sin(x1) + 7*sin(x2)^2 + 0.1*x3^4*sin(x1)
    let indices = estimate_saltelli2010(&saltelli, |x| {
        x[0].sin() + 7.0 * x[1].sin().powi(2) + 0.1 * x[2].powi(4) * x[0].sin()
    });

    println!("{indices}");
}

Methods

Method	Function	Reference
Variance-based (Sobol')
Saltelli 2010	`estimate_saltelli2010`	Saltelli et al. (2010) Comp. Phys. Comm. 181(2)
Jansen	`estimate_jansen`	Jansen (1999) Comp. Phys. Comm. 117(1-2)
Janon	`estimate_janon`	Janon et al. (2014) Math. Comp. Sim. 107
Owen	`estimate_owen`	Owen (2013) ACM Trans. Model. Comp. Sim. 23(1)
Given-data Sobol'	`estimate_given_data_sobol`	Plischke et al. (2013) Eur. J. Oper. Res. 226(3)
Elementary effects
Morris	`estimate_morris_effects`	Morris (1991) Technometrics 33(2)
Grouped Morris	`estimate_grouped_morris_effects`	Campolongo et al. (2007) Env. Mod. Soft. 22(10)
Frequency-based
FAST / eFAST	`estimate_fast`	Cukier et al. (1973); Saltelli et al. (1999)
RBD-FAST	`estimate_rbd_fast`	Tarantola et al. (2006) Rel. Eng. Sys. Safety 91(6)
Distribution-based
Borgonovo delta	`estimate_borgonovo_delta`	Borgonovo (2007) Rel. Eng. Sys. Safety 92(6)
PAWN	`estimate_pawn`	Pianosi et al. (2015) Env. Mod. Soft. 67
QOSA	`estimate_qosa`	Fort et al. (2016) Stat. Comp. 26(1-2)
Derivative-based
DGSM	`estimate_dgsm`	Sobol' & Kucherenko (2009) Math. Comp. Sim. 79(10)
Regression
SRC / SRRC / PCC / PRCC	`estimate_regression_indices`	Saltelli & Marivoet (1990) Comp. Stat. Data Anal. 9(1)
Surrogate
PCE (full OLS)	`fit_full_pce`	Xiu & Karniadakis (2002) SIAM J. Sci. Comp. 24(2)
PCE (sparse LARS/OMP)	`fit_sparse_pce`	Blatman & Sudret (2011) J. Comp. Phys. 230(6)
HDMR	`estimate_hdmr`	Li et al. (2002) J. Phys. Chem. A 106(37)
Active subspaces	`compute_active_subspace`	Constantine (2015) Active Subspaces, SIAM
Game-theoretic
Shapley effects	`estimate_shapley`	Song et al. (2016) SIAM/ASA J. Unc. Quant. 4(1)
Experimental design
ANOVA (two- and three-way)	`estimate_anova_two_way`	Fisher (1925) Statistical Methods
G-theory (D-study)	`estimate_g_theory_pir`	Brennan (2001) Generalizability Theory, Springer
Discrepancy (L2-star)	`compute_discrepancy`	Hickernell (1998) in Monte Carlo and Quasi-Monte Carlo Methods
Fractional factorial	`estimate_fractional_factorial`	Box et al. (1978) Statistics for Experimenters

Crate structure

salib is a facade that re-exports subcrates. Use it for convenience, or depend on individual crates for finer control.

Crate	Role
`salib-core`	`Problem`, `Factor`, `Distribution`, `RngState`, deterministic reductions
`salib-samplers`	LHS, Sobol' QMC, Saltelli, Morris trajectories, FAST/eFAST, Iman-Conover
`salib-estimators`	All estimators listed above
`salib-surrogate`	PCE, sparse PCE, active subspaces
`salib-shapley`	Shapley effects
`salib-validation`	Ishigami, Sobol' G, Morris test functions with closed-form indices
`salib-cli`	`salib sample`, `salib run`, `salib analyze`

Feature flags

Flag	Default	Effect
`samplers`	yes	Sampling designs (LHS, Sobol', Saltelli, Morris, FAST)
`estimators`	yes	All sensitivity estimators
`parallel`	yes	Rayon-based parallel reductions
`surrogate`	no	PCE, HDMR, active subspaces
`shapley`	no	Shapley effects
`validation`	no	Analytic test functions
`serde`	no	`Serialize`/`Deserialize` on all result types
`arrow`	no	`RecordBatch` conversions for Arrow interop
`polars`	no	`DataFrame` conversions (implies `arrow`)
`full`	no	Everything except `serde`, `arrow`, `polars`

Bit-determinism

Identical RngState seeds produce identical results regardless of thread count. Parallel reductions use tree-structured accumulation to eliminate float-associativity nondeterminism under rayon. Disable parallel for serial-only builds; results remain identical.

Citation

If you use salib in published research, please cite:

@software{salib_rs,
  author  = {{antimeme.ai}},
  title   = {salib: Global Sensitivity Analysis for Rust},
  url     = {https://github.com/antimeme-ai/salib},
  version = {0.1.1},
  year    = {2026}
}

For individual methods, see the references in the method table above.

MSRV

1.87

License

MIT OR Apache-2.0, at your option.

salib 0.1.1