Hotelling’s Charts Using Winsorized Modified One Step M-estimator for Individual Non Normal Data

Hotelling T control chart is one of the most important chart used to statistical quality control in the manufacture processes. However, this chart is sensitive to the non normal data and therefore, must be modified to improve its behavior with this kind of data. In this paper, we propose two new alternatives charts based on change the usual mean and covariance estimator by robust location and scale matrix. Thus, we use the winsorized mean and winsorized covariance matrix, respectively. Concretely, the robust scale estimators with highest breakdown points namely MADn and Sn are used to suit the criterion in the modified one step M-estimator (MOM). The control limits for these robust charts are calculated based on simulated data and the assessment of these new alternatives charts is based on the false alarm and the probability of detection out of control observation with non normal data. The results show in general that the performance of the alternatives robust Hotelling’s charts are better than the performance of the traditional Hotelling’s chart.