sobol

sobol(model, data, N=None, nprng=None)

Total and first order Sobol sensitivity indices. https://en.wikipedia.org/wiki/Variance-based_sensitivity_analysis.

Parameters:

• model (function) – Maps data to scores

• data (ndarray) – Data matrix. Each row is a sample vector.

• N (int) – sample size for monte carlo estimate of sobol indices. Should be less than or equal to the number of rows of data. If None, entire dataset is used.

• nprng (RandomState) – Optional numpy RandomState.

Returns:

Total and first order Sobol sensitivity indices. Each index is expressed as an array of length equal to the number of features in the supplied data matrix.