DataTerminology.comNEW
Study of how variations in model inputs contribute to variability in model outputs.
DESCRIPTION
Sensitivity analysis quantifies the influence of input factors on a model’s response. Local methods assess effects near a nominal point via derivatives; global methods explore the full input space using variance-based indices (e.g., Sobol). It guides feature prioritization, robustness checks, and resource allocation. In AI, it relates to explainability and model risk assessment. Proper design requires sampling strategies, convergence checks, and interaction effect consideration. Combined with uncertainty analysis, it informs decision-making under uncertainty.
ALTERNATIVE DEFINITIONS
Wikipedia
Sensitivity analysis is the study of how the uncertainty in the output of a mathematical model can be apportioned to different sources of uncertainty in its inputs.
Canonical definition coupling sensitivity with uncertainty attribution.
NIST Engineering Statistics Handbook
Sensitivity analysis determines how the variation in the output of a model (numerical or otherwise) can be attributed to variations in its inputs.
Statistical engineering perspective tied to practical modeling.
European Commission JRC
Sensitivity analysis is the study of how the uncertainty in the output of a model is apportioned to different sources of uncertainty in the model input.
Policy and scientific guidance emphasizing best practice.
RELATED TERMS
Sobol Indices Global Sensitivity Local Sensitivity Partial Derivative Monte Carlo Data Scenario Analysis Robustness Elasticity Variance Decomposition
COMPETING TERMS
Uncertainty Analysis
KNOWLEDGE AREAS
Mathematics Data Science Testing & Validation MLOps
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Data Analyst
EXTERNAL SOURCES
- Wikipedia - Sensitivity analysis
- SALib - Python library for global sensitivity.
- NIST - Handbook section on sensitivity.
- European Commission JRC - Guidelines on sensitivity and uncertainty.
TOP QUESTIONS
What is the difference between local and global Sensitivity Analysis?
Local methods analyze small perturbations near a baseline; global methods explore the full input space, capturing nonlinearities and interactions.
Category: Comparison
Why perform Sensitivity Analysis in machine learning?
It helps understand feature importance, robustness, and risk. It complements explainability and informs data collection priorities.
Category: Relevancy
What are Sobol indices in Sensitivity Analysis?
They decompose output variance into contributions from inputs and interactions, providing global sensitivity measures.
Category: Concept
What sampling strategies are used in Sensitivity Analysis?
Latin Hypercube Sampling, quasi-random sequences (Sobol, Halton), and Monte Carlo ensure coverage and variance reduction.
Category: Practice
What are common mistakes in Sensitivity Analysis?
Too few samples, ignoring interactions, and conflating correlation with causation lead to misleading conclusions.
Category: Mistakes
How does Sensitivity Analysis relate to Uncertainty Analysis?
They are complementary: uncertainty quantifies output spread; sensitivity attributes that spread to inputs.
Category: Concept
Can Sensitivity Analysis be applied to black-box models?
Yes, global variance-based or derivative-free methods work with black-boxes as long as you can sample inputs and evaluate outputs.
Category: Applications
How many samples do I need for robust Sensitivity Analysis?
It depends on input dimensionality and method; convergence checks and pilot runs guide sample size.
Category: Requirements
What tools support Sensitivity Analysis?
SALib, UQLab, and proprietary tools offer variance-based and screening methods with reproducible workflows.
Category: Tools