A Comparison of Two Tests for the Significance of a Mean Vector.
ROCHESTER UNIV NY DEPT OF STATISTICS
Pagination or Media Count:
Hotellings T-square procedure for testing the significance of a mean vector treats all variables symmetrically and may not be appropriate if the variables carry unequal importance. The step-down procedure, a possibly appropriate alternative, is studied in this paper. An underlying invariance structure is established and then used to develop a canonical form for studying the power functions of the two methods. A Monte Carlo experiment is conducted in this framework and conclusions are reported.
- Statistics and Probability