Change-Point Detection in Multivariate Categorical Processes with Variable Selection

Statistical process control (SPC) has been widely used to control and improve the quality of products in manufacturing processes. Currently, a limited number of schemes is available for change-point detection in multivariate categorical processes (MCPs), where each quality characteristic of products is measured by several attribute levels. Furthermore, existing few methods neglect the dependence structure among quality characteristics so as to provide low efficiency in detecting change points. This paper develops one change-point detection scheme based on the hierarchical log-linear model, which integrates a variable selection procedure to focus only on significant interaction effects, or equivalently dependent relationships among the factors. Therefore, the scheme can achieve higher efficiency than its competing method. The simulation study demonstrates the superiority of the proposed scheme and a real application shows the implementation of the scheme.