Using stepwise discriminant analysis as a post-hoc procedure to a significant Hotelling's T2 control chart

In statistical process control, multivariate control charts are informative tools to a complex process. When the multivariate control chart is deemed to be operating out-of-control, it is important to accurately determine which monitored variable(s) are causing the out-of-control signal. The present study analyzed the effectiveness of stepwise discriminant analysis as a follow-up procedure to a significant Hotelling's T2 control chart. It was argued that given the multivariate nature of the data, it is appropriate to use a multivariate procedure for the follow-up analysis. Results demonstrated that the number of monitored variables, effect size, sample size, and alpha level all affected performance of stepwise discriminant analysis. Overall, small numbers of monitored variables, paired with small alpha levels, large effect sizes, small m, and large n were most successful.