A0421
Title: Functional ANOVA based on Fourier transform of distribution
Authors: Daniel Hlubinka - Univerzita Karlova (Czech Republic) [presenting]
Zdenek Hlavka - Charles University (Czech Republic)
Abstract: Functional two-sample tests based on empirical characteristic functionals are studied. A Cramer-von Mises type test statistic is considered with integration over a preselected family of probability measures, say Q, leading to a computationally feasible and powerful test statistic. Small sample properties of the resulting two- and k-sample functional tests are investigated in a simulation study. In particular, the resulting tests are compared to previously proposed tests of equality of mean functions and covariance operators, and it is shown that a proper choice of the probability measure Q gives very good power in detecting shift and scale alternatives.
