The effect of the size of the training set and number of principal components on the false alarm rate in statistical process monitoring

This paper describes the sensitivity of false alarm rate to misspecification of the number of PCA components in multivariate statistical process control (MSPC) models. Using simulations, it is shown that choosing an incorrect number of components in monitoring models may seriously affect the false alarm rates of the control charts. Furthermore, the false alarm rate becomes worrysome when the size of the training set is small. Using a leave-one-out procedure for building the control charts partly solves this problem.