New multivariate generalized likelihood ratio control charts enhanced with Klein’s sensitizing rules significantly outperform classical monitoring schemes when detecting small shifts in both process mean and dispersion under measurement error.
The study directly addresses recognized operational limitations in conventional statistical quality control methodologies, specifically focusing on traditional Shewhart-type control charts. Although Shewhart-type charts remain widely recognized for their computational simplicity and deliver satisfactory performance when identifying large process shifts, the authors observe that they fail to exhibit sufficient sensitivity when encountering small process changes.
Compounding this sensitivity challenge, real-world manufacturing and engineering applications frequently experience assignable causes that disrupt multiple operational parameters simultaneously. In many industrial environments, an assignable cause alters both the location and the dispersion parameters of a monitored process at the same moment, requiring monitoring frameworks capable of joint parameter evaluation. Furthermore, the researchers emphasize that despite recent technological advances in automated data-collection systems, measurement errors remain practically unavoidable during data gathering, making it critical to explicitly evaluate how such measurement inaccuracies degrade control-chart performance.
Motivated by these operational realities, the research team pursued two distinct developmental objectives to advance multivariate statistical monitoring. First, the authors engineered two enhanced versions of the multivariate generalized likelihood ratio (MGLR) control chart by embedding Klein’s 2-of-2 and 2-of-3 sensitizing run rules, aiming to accelerate the rapid detection of small shifts occurring in process mean, process dispersion, or both parameters concurrently. Second, the investigators adapted these enhanced MGLR charts under an additive covariate model, establishing a formal mathematical structure to account directly for the confounding presence and impact of measurement error.
To evaluate the practical effectiveness of the proposed monitoring schemes and measure the adverse consequences of measurement error, the authors conducted extensive computational simulation experiments. The simulation framework examined the proposed charts across diverse out-of-control operational scenarios alongside multiple levels of error variance. The performance of each chart was quantitatively assessed using three established statistical run-length criteria: the average run length (ARL), the standard deviation of the run length (SDRL), and the median run length (MRL), providing a comprehensive assessment of detection speed and variability.
In addition to baseline performance comparisons, the investigation included an extensive sensitivity analysis exploring the impact of subgroup sample size (n) and process dimensionality (p). The findings demonstrated that increasing the subgroup sample size substantially improves the shift detectability of all evaluated control charts, allowing faster identification of out-of-control conditions. Conversely, the sensitivity analysis revealed that process dimensional growth noticeably degrades detection sensitivity across the charts, highlighting the complicating effect that higher dimensional data spaces exert on multivariate likelihood ratio surveillance.
The study’s comparative results confirmed that measurement error diminishes the detection sensitivity of all evaluated control charts, with performance degradation clearly observable even at low levels of error variance. Nevertheless, both the 2-of-2 and 2-of-3 sensitizing schemes substantially outperformed the classical MGLR chart across all examined shift scenarios and error variance levels. Among the improved designs, the 2-of-3 sensitizing rule achieved slightly superior detection capability compared to the 2-of-2 scheme, confirming the value of sensitizing rules for robust joint process monitoring.
Subject of Research: Statistical process monitoring using multivariate control charts under measurement error
Article Title: Improved MGLR control charts based on Klein’s sensitizing rules for joint mean and dispersion monitoring under measurement error
Article References: Maleki, M. R., Jafari, M., Eghbali, H., & Nikoogoftar, M. (2026). Improved MGLR control charts based on Klein’s sensitizing rules for joint mean and dispersion monitoring under measurement error. Scientific Reports. https://doi.org/10.1038/s41598-026-81124-6
Image Credits: AI Generated
DOI: 10.1038/s41598-026-81124-6
Keywords: multivariate generalized likelihood ratio, MGLR control chart, Klein's sensitizing rules, 2-of-2 sensitizing rule, 2-of-3 sensitizing rule, measurement error, additive covariate model, average run length, standard deviation of run length, median run length, subgroup size, process dimensionality, statistical process monitoring, Shewhart control charts, joint parameter monitoring, dispersion shifts, mean shifts, error variance
Cite Scienmag News
Denise Maddox. (November 19, 2026). Klein’s Sensitizing Rules Improve Multivariate Control Chart Performance Under Measurement Error. Scienmag. https://scienmag.com/kleins-sensitizing-rules-improve-multivariate-control-chart-performance-under-measurement-error/
Denise Maddox. "Klein’s Sensitizing Rules Improve Multivariate Control Chart Performance Under Measurement Error." Scienmag, 19 November 2026, https://scienmag.com/kleins-sensitizing-rules-improve-multivariate-control-chart-performance-under-measurement-error/. Accessed 19 November 2026.
Denise Maddox. "Klein’s Sensitizing Rules Improve Multivariate Control Chart Performance Under Measurement Error." Scienmag. November 19, 2026. https://scienmag.com/kleins-sensitizing-rules-improve-multivariate-control-chart-performance-under-measurement-error/








