An improvement of the mcd robust hotelling's T2 through control limits

The hotelling t2 control chart has been widely used for monitoring several quality characteristics measured simultaneously in a product or process. the presence of multiple outliers in phase-i is not detected by this chart,due to the masking effect. this is why 2 robust alternatives to hotelling's t have been developed. these are based on several robust estimators,although the most commonly used one is based upon minimum covariance determinant (mcd) estimators. 2 as the distribution of the t statistic based on mcd estimators is unknown in closed form,in order to implement this control chart,control limits approximated to several distributions are usually used or generated by means of a simulation procedure. in the present paper,we use a monte carlo simulation and real examples to evaluate the performance of this robust control chart when different control limits are used. moreover,we propose a new way to determine the control limits with an adjusted function for different samples sizes and dimensions of the process. © 2013 pakistan journal of statistics.