A Nonparametric Adaptive EWMA Control Chart for Binary Monitoring of Multiple Stream Processes

A team of researchers including Faruk Muritala, Austin Brown, Dhrubajyoti Ghosh, and Sherry Ni has published a new statistical method called the Cumulative Standardized Binomial EWMA (CSB-EWMA) control chart. This tool addresses a critical gap in Statistical Process Control (SPC) for monitoring binary proportions across multiple independent data streams, with applications ranging from manufacturing quality control to cybersecurity threat detection. Unlike traditional EWMA charts that rely on asymptotic variance approximations—which fail during early-phase monitoring—the CSB-EWMA derives the exact time-varying variance of the EWMA statistic. This enables adaptive control limits that maintain statistical rigor from the very first sample, eliminating the approximation errors that plague existing methods.

Through extensive simulations, the researchers identified optimal smoothing (λ) and limit (L) parameters to achieve target in-control average run lengths (ARL0) of 370 and 500 samples. The chart demonstrates exceptional performance in detecting process shifts, with out-of-control average run lengths (ARL1) dropping to just 3-7 samples for moderate shifts (δ=0.2). This represents detection speeds approximately 50 times faster than the in-control baseline. The method also shows remarkable robustness across different data distributions, maintaining low ARL1 Coefficients of Variation (CV < 0.10 for small shifts) for both ARL0 targets. The CSB-EWMA provides practitioners with a distribution-free, sensitive tool that doesn't require normal distribution assumptions, making it particularly valuable for real-world applications where data often violates traditional statistical assumptions.