Robust adaptive EWMA–CUSUM control charts for monitoring autocorrelated and non-normal processes

Control charts are essential for monitoring process stability; However, traditional methods fail when data exhibit both autocorrelation and non-normality. This study introduces a Robust Adaptive EWMA–CUSUM (RAEC) chart that integrates robust estimators, adaptive shift estimation, and explicit accommodation of serial dependence. Using 130 daily observations of the KESONIA interest rate benchmark from Kenya’s Central Bank (September 2025–March 2026), a period containing known structural breaks, the methodology is evaluated through extensive Monte Carlo simulations under normal, heavy-tailed t(3), skewed lognormal, and contaminated distributions. The RAEC chart achieves substantial improvements over existing methods, with gains of 82.3% over classic CUSUM and 96.0% over classic EWMA for small shifts (δ = 0.2). It detected emerging shifts approximately 45 days before the first policy announcement, demonstrating practical utility for financial stability monitoring. Bootstrap uncertainty quantification, sensitivity analyses, and benchmark comparisons confirm the framework’s robustness and superior detection capability, providing practitioners with a powerful tool for reliable process surveillance under challenging conditions.