2 Methods
In this section, we discussed on the modified Ft method,
which combines Ft statistics with scale estimators
suggested by [14].
2.1 Ft Statistics
The original Ft statistic or trimmed F statistic was
introduced by [8]. This statistical procedure is able to
handle problems with sample locations when non
normality occurs but the assumption of homogeneity of
variances still applies. This new statistic is easy to
compute and is used as an alternative to the classical F
method involving one-way independent group design.
To further understand the Ft method, let
X(1) j;X(2) j; : : : ;X(nj) j
be an ordered sample of group j with size nj.
We calculate the trimmed mean of group j by using:
Xt j =
1
nj −g1 j −g2 j
[
nj−g2 j
å
i=gi j+1
X(i) j]
where
= number of observations X(i) j such that gi j that
(X(i) j − ˆM j) < −2:24(scale estimator),
where
number of observations X(i) j such that g2 j that
(X(i) j − ˆM j) > 2:24(scale estimator),
ˆM
j = median of group j, and
scale estimator = MADn, Tn or LMSn.
2 MethodsIn this section, we discussed on the modified Ft method,which combines Ft statistics with scale estimatorssuggested by [14].2.1 Ft StatisticsThe original Ft statistic or trimmed F statistic wasintroduced by [8]. This statistical procedure is able tohandle problems with sample locations when nonnormality occurs but the assumption of homogeneity ofvariances still applies. This new statistic is easy tocompute and is used as an alternative to the classical Fmethod involving one-way independent group design.To further understand the Ft method, letX(1) j;X(2) j; : : : ;X(nj) jbe an ordered sample of group j with size nj.We calculate the trimmed mean of group j by using:Xt j =1nj −g1 j −g2 j[nj−g2 jåi=gi j+1X(i) j]where= number of observations X(i) j such that gi j that(X(i) j − ˆM j) < −2:24(scale estimator),wherenumber of observations X(i) j such that g2 j that(X(i) j − ˆM j) > 2:24(scale estimator),ˆMj = median of group j, andscale estimator = MADn, Tn or LMSn.
การแปล กรุณารอสักครู่..