Uncovering heterogeneous effects in computational models for sustainable decision-making

This paper formalizes heterogeneous effects in computational models and introduces a simple binning algorithm for detecting the most important first- and second-order effects directly from a single Monte Carlo dataset.

By coupling variance-based sensitivity analysis with Simulation Decomposition (SimDec), the procedure makes SimDec fully automatic: the algorithm identifies influential inputs, detects interactions, and selects variables for decomposition without trial-and-error.

The framework is demonstrated on environmental, financial, and sequential decision-making models, and is implemented in open-source packages across Python, R, Julia, and Matlab.