An Extension of MIC for Multivariate Correlation Analysis Based on Interaction Information

Maximal Information Coefficient (MIC) is a novel bivariate correlation analysis method that attracts a lot of attention from various research fields, due to its generality and equitability. So on the basis of interaction information, we propose to extend MIC into the context of multivariate correlation analysis. From the results of simulative experiments, our propose method can detect a wide range of associations by assigning 1 to noiseless functional data sets and 0 to statistically independent data sets, verifying its generality, equitability and robustness.