Authors : Ayu Abdul Rahman, Sharipah Soaad Syed Yahaya, Abdu Mohammed Ali Atta, Nor Aishah Ahad and Hashibah Hamid
Abstract: In applying control chart with estimated parameters for monitoring changes in a process, Phase I samples are typically assumed to be free of outliers or any other data anomaly. Naturally, the sample mean and the sample standard deviations are used as estimators, yielding efficient estimates for the chart. Nonetheless, when Phase I may be contaminated, this regular practice is no longer suitable as classical estimators are susceptible to the effect of outliers which in turn may affect control chart performance. This study shows that the effect is not trivial via. the application of EWMA control chart. Moreover, this study focuses on the effect using alternative and robust Phase I estimators on the EWMA when the chart is used to monitor changes in the process mean. In this study, an automatic trimmed mean estimator is used to provide estimate for the process mean. Meanwhile, for the standard deviation of the process, this study employs three different estimators including the corresponding robust scale estimator used in the trimming process of the location measure. Simulated data were used to test the performance of the EWMA control charts. The finding based on mean and percentiles of the run-length distribution shows quicker detection of out-of-control status when robust statistics were used to compute parameter estimates in Phase I of the EWMA chart upon contamination in the data set.
Ayu Abdul Rahman, Sharipah Soaad Syed Yahaya, Abdu Mohammed Ali Atta, Nor Aishah Ahad and Hashibah Hamid, 2020. The Impact of Data Anomaly on EWMA Phase II Performance. Journal of Engineering and Applied Sciences, 15: 3010-3015.