When a T2 control chart produces an out-of-control signal, it gives no direct information concerning which subset of the variables is responsible. A new method is presented for addressing this problem. It uses the "difference in beta" statistic (DFBETA), which measures the influence of a single observation on the given regression coefficients. The motivation for using the DFBETA is that if a regression coefficient will be significantly changed with the exclusion of a specific case, the variable associated with that DFBETA may be out-of-control. For processes with certain types of variance-covariance structures, the method presented is better at detecting single variable shifts than many existing methods. |