Polynomial Chaos Expansion (PCE) is one of the most popular surrogate modeling algorithms, especially for global sensitivity analysis. Using PCE, the first-order and the total Sobol' indices and univariate effects are obtained analytically. Harenberg et al. (2019) is the first example in economics to present global sensitivity analysis based on PCE.
We study PCE using the chaospy
library and apply the technique to some test functions. In python
, there are some implementations of the PCE-based global sensitivity analysis, such as uqpylab
. We verify our implementations by comparing them with the results from uqpylab
.