Outliers’ detection with a sensitive exponentially weighted moving average control chart

Abstract

The effect of parameter estimation emanating from the retrospective stage on the monitoring stage of control charts cannot be overemphasized. These effects are born of but are not limited to the practitioner-to-practitioner variations in the amount and type of samples employed to estimate the process parameters. Another major factor to this effect is outlying errors in phase-I data. This study evaluates the performance of the exponentially weighted moving average (EWMA) control chart, based on outlying values and practitioner-to-practitioner's variation in the phase-I preliminary samples. Furthermore, the study proposes a sensitive EWMA control chart with Tukey's and median absolute deviation (MAD) outlier detectors. We study the proposed EWMA chart's estimation effect based on the outlier detector models compared to the default EWMA chart through the Monte-Carlo simulation approach. By studying the run length properties of the proposed schemes, the study's findings prove that the outlier detectors-based models are more stable in the presence of outliers and require less observation in the retrospective stage. The study concludes by implementing the results on a real-life dataset extracted from the semiconductor manufacturing industry.