Monitoring multivariate coefficient of variation with individual observations

Abstract

AbstractMonitoring the multivariate coefficient of variation (MCV) is conducted when the multivariate process variation changes proportionally with the mean vector because under this situation the traditional multivariate charts for the mean and variance cannot work correctly. All existing MCV charts in the literature are implemented based on sample MCVs computed from samples of size larger than unity each. However, there are situations when forming samples (of ) is not feasible. To address the aforementioned limitation of existing MCV charts, in this paper, an EWMA chart based on individual observations for monitoring the MCV is developed. Numerical studies based on Monte Carlo simulations are conducted to investigate the run length properties of the proposed chart in detecting either increases or decreases in the MCV, for two and three quality characteristics and various parameter values. A real bivariate dataset on the diameter and length of dowel pins is adopted to illustrate the implementation of the proposed chart.