sensitivityanalyser

Sensitivity analyser is a tool or approach used to quantify how the outputs of a model respond to changes in its input parameters. It helps identify which parameters drive results, assess the robustness of conclusions, and guide data collection and model refinement. The term is used in many fields and may appear as analyser (British English) or analyzer (American English).

Methods range from local sensitivity analysis, which estimates the effect of small perturbations around a baseline point, to global sensitivity analysis, which probes the full range of parameter uncertainty. Global methods include variance-based techniques such as Sobol indices and screening methods like the Morris approach; FAST and related algorithms are also common. For expensive models, emulators or Monte Carlo simulations are often used.

A typical workflow involves defining the model and parameter uncertainties, selecting a sensitivity method, running simulations or evaluations, computing sensitivity metrics, and interpreting results to inform model development or decision making.

Applications span environmental and civil engineering, pharmacometrics, economics, risk assessment, and software systems. Limitations include dependence on assumed uncertainty distributions and parameter correlations, potential nonlinearity, and computational cost. Sensitivity analysis is most informative when paired with uncertainty quantification and model validation.