A Note On A Multivariate Distribution-free Control Chart Based On Log-Linear Model

Statistical process control (SPC) is widely applied to the detection of various production processes. Today, the simultaneous detection of multiple quality characteristics has become very common. A nonparametric multivariate cumulative sum (CUSUM) procedure was proposed by Qiu based on the Log-linear model for detecting shifts in the location parameter vector of the measurement distribution. In this paper, a further investigation is given when the partitioning intervals is more than two and the performance of the chart is measured in terms of ARL and SDRL. The Person test statistic is replaced by the likelihood ratio statistic (LRT). By comparing the simulation, the method can reduce the variance. The EWMA procedure is introduced to design the control chart in order to further improve the chart’s performance.