ROBUSTIFICATION OF SHEWHART CONTROL CHART BY MEDIAN BASED ESTIMATORS: A STUDY ON MALAYSIA STOCK DATA

Statistical control chart is vastly used in financial field, especially in assessing changes in stock returns. Control charts are becoming an important consitituent of the decision-making process in stock trading. However, in the presence of non-normality, the classical control charts may produce too many false trade signals and therefore, may no longer be reliable in stock trading. This study aims to construct robust control chart using median based estimators and subsequently, used it on real financial data. Median and trimean estimators were applied in the construction of the limits for the Shewhart chart as well as in computing the charting statistics. The practical application of the proposed robust charts were demonstrated using a real data set about Top Glove stock's open price. In analyzing the capabilities of the robust Shehwhart charts for stock trading against the classical Shewhart chart, the findings show that the classical chart yields too many trade signals which most are considered as false alarms unlike the robust charts. This paper shows that the robustification of the Shewhart structure via the median based estimators help to alleviate the impact of non-normality on the chart's performance.