Toward SALib 2.0: Advancing the accessibility and interpretability of global sensitivity analyses

Abstract

Sensitivity analysis is now considered a standard practice in environmental modeling. Several open-source libraries, such as the Sensitivity Analysis Library (SALib), have been published in the recent past aimed at simplifying the application of sensitivity analyses. Still, there remain issues in software usability and accessibility, as well as a lack of guidance in the interpretation of sensitivity analysis results. This paper describes the changes made and planned to SALib to advance the ease with which modelers may conduct sensitivity analysis and interpret results. We further offer our perspectives from the past 7 years of maintaining SALib for the consideration of those aspiring to launch their own software for sensitivity analysis, develop methodology, or those otherwise interested in becoming involved in a project like SALib. These include the value of a community of practice to foster best practices for sensitivity analysis, the potential for collaboration across different software (for sensitivity analysis) platforms, and the need to specifically support the software development that underpins computational science.

A cell-based framework for modeling cardiac mechanics

“…The sensitivity analysis was performed using the Python library SALib (Herman and Usher 2017 ; Iwanaga et al. 2022 ).…”

“…This analysis was done for all deformation modes used in the stretching and shear experiments, tracking load values of interest, as well as for averaged stresses across the entire domain under fiber direction stretch experiment FF and contraction. The sensitivity analysis was performed using the Python library SALib (Herman and Usher 2017;Iwanaga et al 2022).…”

Evaluating Implementation Uncertainties and Defining Safe Operating Spaces for Deeply Uncertain Cooperative Multi‐City Water Supply Investment Pathways

“…This global sensitivity analysis method compares the entire probability distribution of both input and output parameters to estimate the sensitivity of the output to a specific input parameter. This study implements Delta Moment‐Independent sensitivity analysis using the SALib Python package (Herman & Usher, 2017; Iwanaga et al., 2022) to identify candidate actions whose perturbations are most likely to change the region and its member utilities' performance and robustness. Another advantage of the Delta Moment‐Independent sensitivity analysis method is its ability to exploit our existing implementation tolerance sampling illustrated in Figure 4 and summarized in Section 3.3.…”

Bayesian quantification of strongly-interacting matter with color glass condensate initial conditions

“…This analysis is an example of analysis of variance (ANOVA) that decompose the total variance of a model into variance of model parameters (at first order), pairs of model parameters (second order), and so on. The first-order Sobol indices quantify the global variance in model observable due to variance in model parameters [128] and are readily available [129,130].…”