A1560
Title: Testing independence in high-dimensional data based on the absolute cross-covariance sum
Authors: Johan Lim - Seoul National University (Korea, South) [presenting]
Seonghun Cho - Inha University (Korea, South)
Seongoh Park - Sungshin Women\'s University (Korea, South)
Abstract: To test the independence of a variable of interest and all other variables, the absolute cross-covariance sum (ACS) statistic is proposed, defined as the sum of absolute correlations. Asymptotic normality of this sum is established as the number of variables increases under regularity conditions and finite-dependence assumptions. Consistent estimators of its mean and variance are proposed, such that the standardized ACS statistic achieves asymptotic normality. Numerical investigations of the size and power of the proposed test statistic are conducted and compared with existing methods. The ACS statistic is applied to identifying island variables, which are independent of others, in gene expression data.
