F. Castro-Prado, D. Edelmann, J. J. Goeman
Distance covariance is a general dependence measure that can be defined on general semimetric spaces, which can be shown to be dual to the Hilbert–Schmidt independence criterion popular in the machine learning community. Both concepts, in turn, present some interesting dualities with the "global tests" of Gaussian. We present some novel tests for the association of single nucleotide polymorphisms (SNPs) with quantitative responses, using this unified approach. We show that certain versions of distance covariance correspond to locally most powerful tests for specific statistical models leading to insights in which situations these tests perform well, which are of paramount interest in the application fields. Closed form expressions for the distributions of the test statistics and corresponding estimators are obtained from the spectral decomposition of the corresponding operators. The performance is illustrated via simulations and real data.
Keywords: Distance covariance, genomics, biostatistics, big data
Scheduled
Biostatistics I
June 9, 2022 10:10 AM
A13
