Joint Research Conference
June 24-26, 2014
Most of the existing multivariate control charts were developed under a specific distributional assumption, which typically has been the multivariate normality assumption. In contrast, nonparametric multivariate control charts do not require specification of a particular family of multivariate distributions, and therefore are more desirable in practice. In this talk, we will first describe two new nonparametric multivariate CUSUM procedures for detecting location and scale changes. These two procedures can be considered as the nonparametric counterparts of the two parametric multivariate CUSUM procedures developed in Crosier (1988). Similar to most existing nonparametric multivariate control charts, these two control charts involve some tuning parameter, which needs to be pre-specified to implement those control charts. The choice of the tuning parameter affects the detection power of the resulting control chart. To choose the appropriate tuning parameter to achieve optimal performance, it usually requires the information about the out-of-control distribution. However, in practice, it is rarely known in advance what the out-of-control distribution is. To overcome this limitation, in the second part of my talk I will introduce a new nonparametric multivariate control chart using a hypothesis testing-based approach. The proposed control chart does not depend on any tuning parameter, and can be considered as a natural generalization of the generalized likelihood ratio chart to the nonparametric setting. We will demonstrate the performance of our proposed control charts using both simulated data and a real manufacturing process data.