Robust procedures for multi-sample location problems with unequal group variances

Abstract

Robust procedures are proposed for testing the equality of several group means without assuming the equality of group variances. These statistics are obtained by modifying Welch's W and Brown-Forsythe's F^* using a trimmed mean and a sine-wave M estimator.Approximate distributions of these new statistics are obtained under normality. Their performances are evaluated by Monte Carlo sampling experiments under various long-tailed symmetric distributions

Keywords:

• Robust procedures
• multi-sample location problem
• heteroscedasticity
• long-tailed distributions
• trimming
• M estimator

^†Research supported in part by the Natural Sciences and Engineering Research Council, Canada.

^†Research supported in part by the Natural Sciences and Engineering Research Council, Canada.

Notes

^†Research supported in part by the Natural Sciences and Engineering Research Council, Canada.