desired is an inference procedure, which in some sense
does almost as well as if the assumption is true, but does
not perform much worse within a range of alternatives to
the assumption. The theories of robustness consider
neighborhoods of parametric models and thus belong to
parametric statistics. A robust procedure usually adopts
what might be called an ”applied parametric viewpoint”,
which according to [5] uses a parametric model. This
model is hopefully a good approximation to the true
underlying situation, but we cannot assume that it is
exactly correct. Frequently in discussions of robustness,
the assumed distribution (probability density function) is
normal; therefore, the type of robustness of interest is
”robustness to non-normality”.
The proposed procedure of modified Ft to be adopted
in this study is among the latest procedures in robust
statistics, was proposed by [10]. This procedure is for
testing the equality of the central tendency measures for J
groups with H0 : q1 = q2 = : : : = qJ , where qJ is the
central tendency parameter corresponding to distribution
FJ : J = 1;2; : : : ;J. Modified Ft uses trimmed mean as the
central tendency measures.