Sample-based estimation of correlation ratio with polynomial approximation

Abstract
Sensitivity analysis has become a natural step in the uncertainty analysis framework. As there is no general sensitivity measure that would capture all information on impact of input factors on model output, analysts tend to combine various measures to obtain a broader image of interactions between different modes. This article concentrates on the correlation ratio, demonstrates methods for calculating this quantity efficiently and accurately, and compares the results. A new method inspired by artificial intelligence techniques emerges as outperforming the familiar methods.
Funding Information
  • Defense Advanced Research Projects Agency (MDA972-02-C-0038)

This publication has 12 references indexed in Scilit: