On the Robustness of Rank Order Tests and Estimates in the Generalized Multivariate One-Sample Location Problem.
INDIANA UNIV BLOOMINGTON DEPT OF MATHEMATICS
Pagination or Media Count:
For stochastic vectors, not necessarily identically distributed, the sign invariant permutation distribution theory of rank order statistics is developed and utilized in the construction of a class of conditionally distribution-free tests for diagonal symmetry. The asymptotic multinormality of the rank order statistics, when the null hypothesis of sign-invariance is not necessarily true, is extended to the case when each observation may have different distribution. Some interesting inequalities on the covariance matrix of rank order statistics are obtained. These results are used in studying the robustness of rank order tests and estimates for heterogeneous distributions. Author
- Statistics and Probability