A Robust Standard Deviation Control Chart Based on Trimmed and Screened S and R‐Based Estimators

ABSTRACT This article proposes robust standard deviation estimators for Phase I analysis in statistical process control. Four new estimators are developed that extend conventional range and sample standard deviation statistics by incorporating trimming and screening mechanisms. The proposed estimators are compared with established robust methods, including the biweight A estimator, M‐estimators, and median‐based approaches. Performance is evaluated via mean squared error under normality and four contamination scenarios: diffuse symmetric, diffuse asymmetric, localized variance, and diffuse mean disturbances at 6% and 10% contamination rates. Simulation results demonstrate that the proposed estimators, particularly those based on trimmed moving range and trimmed standard deviation with dual‐stage screening, consistently outperform alternatives. Phase II control chart performance is assessed through average run length (ARL), expected average run length, and conditional ARL under both known and unknown process shifts. The results show substantial improvements in detection capability and robustness.