A performance comparison of AR(1) Shewhart process monitoring between residuals and raw data

In today’s competitive business environment, ensuring and maintaining quality is of the greatest importance for companies. Monitoring processes is crucial to maintain product and service quality, with quality directly linked to process variability. Statistical Process Control (SPC) helps practitioners to identify two types of variation: common causes inherent to the process and special causes arising from external factors. SPC uses control charts to observe process behavior over time. Traditional control charts assume independence and identical distribution of observations, a condition often unmet in real-world processes, potentially leading to an excess of false alarms. Methods have been developed to adapt control charts when independence is violated, but a comprehensive comparison between using residuals and raw data in AR(1) Shewhart process monitoring is lacking. This thesis aims to bridge this gap by comparing the performance of the two approaches, identifying conditions favoring one method over the other. The study explores different levels of autocorrelation, mean changes, and diverse control limits. The findings enhance the practical application of control charts in scenarios challenging the assumption of independent observations.

https://orcid.org/0000-0002-5196-3451