Introduction

Sensitivity analysis is a cornerstone of data-driven research, helping quantify how input variables affect observed outputs.

Key Concepts

• Sensitivity Analysis: Quantifies the influence of input variables on outputs.
• Data-driven Approach: Operates directly on empirical datasets rather than pre-defined models.
• Applications: Identifying dominant variables, assessing interactions, and improving interpretation of complex datasets.

Local vs. Global Sensitivity Analysis

Sensitivity analysis can be broadly categorized into local and global approaches. Local sensitivity analysis assesses the effect of small perturbations around a single reference point, offering information about the immediate response of the system. While computationally efficient, it provides a limited perspective and often neglects nonlinear effects and interactions between variables. Global sensitivity analysis, on the other hand, considers the entire range of input variability and captures nonlinearities and higher-order interactions, offering a more comprehensive understanding of input-output relationships. This makes global methods particularly suited to complex, multivariate datasets typical in geochemical research.

Toolbox Focus

The Sensitivity Analysis Toolbox is designed for global, data-driven sensitivity analysis, with a structured workflow including:

• Pre-analysis Stage:

  • Sampling / Design of Experiments (DOE) techniques

• Analysis Stage:

  • Correlation / Regression method
  • Variance-based decomposition
  • Quantification of variable contributions and interactions

By integrating these techniques, the toolbox enables researchers to:

• Identify dominant variables
• Detect parameter interactions
• Interpret empirical datasets with higher confidence

This theoretical foundation sets the stage for the subsequent sections of the guide, where sampling methods and global sensitivity analysis techniques are discussed in detail.