Abstract
Multivariate sign-based tests for a class of testing problems on the eigenvalues of scatter matrices are constructed. The class of testing problems is characterized by real mappings $h$ say. A necessary and sufficient condition on $h$ to obtain asymptotically valid sign-based procedures is identified. A simulation study shows the very good robustness properties of our sign tests while their practical relevance is illustrated on a real data set.
Type
Publication
Econometrics and Statistics
