Page 117 - Special Topic Session (STS) - Volume 1

STS410 Abdul Ghapor H. et al.
coding of the programming is developed using Tibco SPLUS statistical
software in assessing the power of performance of FDMCEC statistic in
detecting the outlier.
Step 1: The values of X variable are generated from the von Mises
distribution and in the size of n = 70, 100 and 130 and   10, 15 and 20,
*
respectively. An observation Xd is then contaminated with some levels of
contamination  where the level of the contamination are  = 0.2, 0.4,
0.6, 0.8 and 1, respetively. The formula of contaminating the observation is
as follow:
X d * X i   (mod 2  )
Step 2: Find Y according to the generated X. The variables X and Y are
considered with generated random error terms of  i ~VM , 0 ( )  and
 i ~VM , 0 ( )  , respectively where   .  The variables are fitted to the

unreplicated LFRM with parameter estimation as described Section 4.2.
Step 3: The values of functional mean circular error cosine are calculated
for all observations.
th
Step 4: Omit the i observation of the generated data, where i=1, 2, 3, …,n
to obtain FMCEC(-i). Repeat this step for all i observations to obtain the set
of value FMCEC(-i).
Step 5: Calculate the absolute difference between FMCEC and FMCEC(-i).
Then, find value of FDMCEC  FMCEC  FMCEC  for all i.
( ) i ( ) i
Step 7: Determine the values of FDMCEC(-i) that exceed the cut-off
equations developed in Section 6.3. If they exceed, they are marked as
outliers.
Step 8: Steps 1 to 7 are repeated for 500 simulation and the percentage of
correct outlier detection is calculated as the power of performance. Table
6.7 shows the power of performance of FDMCEC in outlier detection.

106 | I S I W S C 2 0 1 9