Phase I Multivariate Coefficient of Variation Control Charts for High‐Dimensional Processes

ABSTRACT Multivariate control charts have traditionally focused on monitoring the process mean vector and/or covariance matrix. Recent studies have extended this framework to Phase II monitoring of the multivariate coefficient of variation (MCV); yet very limited work has examined MCV‐based monitoring in Phase I. This gap is critical because Phase I procedures establish the IC state by estimating control limits from historical data, which may contain contamination or atypical observations. Robust Phase I methods are therefore essential for identifying and removing such contaminated samples before deploying Phase II monitoring. However, no existing research has evaluated the performance of MCV charts for high‐dimensional Phase I processes, where covariance estimation becomes unstable and classical methods fail. This study investigates the Phase I performance of several MCV control charts using a reduction approach built on the Least Absolute Shrinkage and Selection Operator (LASSO)‐penalized likelihood ratio framework. The proposed charts are evaluated under the localized coefficient of variation disturbance scenarios using the probability to signal () as the primary performance metric. A real‐world case study using diagnosed breast cancer data demonstrates the practical application of the proposed Phase I methodology. The results provide valuable guidance for practitioners in industrial and healthcare environments by identifying efficient and computationally stable Phase I control charts for monitoring the MCV of a process in high‐dimensional environments.