spurious rejections of the null hypotheses and the power
of test can be substantially reduced from theoretical
values, resulting in undetected differences. Realizing the
need for a good statistic in addressing these problems, we
integrate the Ft statistic [8] with the highest breakdown
scale estimators [14] and these new methods are known as
the modified Ft methods.
This paper has shown some improvement in the
statistical solution for detecting differences between
location parameters. The findings showed that the
modified robust procedures, Ft with MADn, Ft with Tn, Ft
with LMSn are comparable with Kruskall Wallis in
controlling Type I error rates under most conditions. In
the analysis on real data, Ft with Tn (p = 0.0020) and Ft
with MADn (p = 0.0021) showed stronger significance
than Kruskall Wallis (p = 0.0160). Even though Ft with
LMSn (p = 0.0407) showed weaker significance than the
aforementioned procedures, its performance was proven
to be much better than the parametric ANOVA (p =
0.0870) in both simulation study and real data analysis.