Multivariate Normality Test for the detection of Weak Tremors
Title : Multivariate Normality Test for the detection of Weak Tremors
Asbtract : In this talk, I will first present a procedure for testing that a multivariate colored stationary time-series is Gaussian. This test is based on the multivariate definition of the Kurtosis initially proposed by Mardia (1970). The complete derivation of the test statistic's distribution under the null hypothesis of Gaussianity is given for the bivariate case and the general multivariate case will be handled using bivariate random projections. I will then proceed to a practical implementation of the normality test, and a study of its performance will be carried for various scenarios on both synthetic data (using colored copulas) and real seismic data.
More information : https://www.gipsa-lab.grenoble-inp.fr/page_pro.php?vid=3524