A simulation-based Bayesian multivariate adaptive EWMA framework with hybrid score functions for monitoring water quality
Fozia Tauqeer, Muhammad Riaz, Babar Zaman, Irshad Ahmad Arshad · Journal of Statistical Computation and Simulation · 2025
A simulation-based performance demonstrated the novelty of integrating a Bayesian multivariate adaptive exponentially weighted moving average (MAEWMA) framework for significant improvement in quality control. The Bayesian MAEWMA control chart is designed to track complex quality profiles in manufacturing and environmental systems. The proposed method enhanced the detection ability of small to moderate shifts in multivariate process control. An adaptive approach improved control chart performance by adjusting parameters. A comprehensive sensitivity analysis has been conducted to examine how the structure of prior distributions and the selection of hyperparameters affect key performance measures, such as Average Run Length (ARL), Standard Deviation of Run Length (SDRL) and Median Run Length (MRL). Additionally, Highest Density Regions (HDRs) have been evaluated using ellipsoidal contours derived from the variance–covariance matrix and the posterior mean vector, which helped to convey uncertainty from prior analyses. Robustness has been assessed by decision-theoretic performance under various loss functions. The effectiveness of the proposed control chart has been established through a case study of water quality monitoring, which examined five related factors: pH, hardness, chloramine, electrical conductivity and total organic carbon.