Sensitivity Analysis as an Ingredient of Modelling.

Inadequacy of existing techniques for sensitivity analysis (SA) of models, coupled with an overconfident attitude on the side of the model developers, has so far played against the use of proper SA. By reviewing recent methodological developments in the field of global quantitative sensitivity analysis, we argue that time is ripe for a wider use of SA. The paper focuses on: I. A general but rigorous definition of SA II. The importance of the use of proper quantitative SA methods III. The tasks where SA can be useful We also ponder a new paradigm for the use of model in experiments, whereby the rules of the game involving a model and the world to which it is applied are explored by quantitative SA techniques before the experiment is enacted, and address the merit of modeling activities in the scientific method.

SALTELLI Andrea;  TARANTOLA Stefano;  CAMPOLONGO Francesca; 

1998-10-13

JRC17166

https://publications.jrc.ec.europa.eu/repository/handle/JRC17166,