The results from the Monte Carlo simulation show the eight
different methods for detecting outliers. The best of Y’s
outliers are (i ) r and i CD methods. This is
important (i ) r perform better than i CD , because (i ) r mainly
show high values of the detection outlier of every the sample
sizes and the percentages of Y’s outliers. The next best
statistics for detection are i d and i r methods. They have good
outlier detection when large sample sizes and high the
percentage of Y’s outliers. The ii h and i t methods have values
of the detection outlier with small sample sizes, but
compromised outlier detection when the large sample size
and the percentages of outliers are increased. The best of X’s
and both X’s and Y’s outliers is i MD method. It has the
highest values of detection outlier when the presence the
sample sizes are small, medium and large sizes. The next best
statistics for the detection are Hat matrix( ) ii h , Cook’s square distance ( )i CD and i DEFFIT . The i DEFFIT method has
more the values of detection outlier when less than outliers.
Although show (i ) r , i CD and i MD methods are clearly
favorable to outlier detection methods, given our methods
success in the identification of outliers. They can also be
considered for use in estimation. One can estimate the
regression coefficients with outliers by applying the robust
regression. The estimation method is applying a down
weighing approach would be worthwhile.