Robustness of dispersion control charts in skewed distributions

This study examines the relative efficiency and the finite sample breakdown point of eight different estimators in Phase I of the control charting process when outliers occur in non-normal data. The performance of control charts based on these estimators is investigated by using average run lengths under four disturbances in three skewed distributions. The simulation result shows that control charts based on the modified biweight A estimator (D7) and the median of the absolute deviations (MAD) from the median are more robust than those in highly skewed distributions. In practice, in addition to robustness, computational simplicity is another important factor for practitioners when they are choosing control charts. It is thus suggested the control chart based on the MAD should be considered first due to its simplicity and robustness.