Abstract: | To test an assumed mean vector, and to test the equality of two mean vectors, robust statistics are developed which have exactly the same form as the Hotelling T2 statistics. These statistics are shown to have remarkable type I error robustness and power. |