Willis A. Jensen, Jeffrey B. Birch, William H. Woodall
Profile monitoring is used when the product or process quality is best represented by a function at each time period. Often some parametric method is used and then the estimated parameters are monitored over time to determine if there have been changes in the profiles. Unlike previous modeling methods that do not incorporate a correlation structure within the profiles, the linear mixed models proposed can monitor the linear profiles to account for any correlation structure within a profile. It is shown that when data are balanced, there is no advantage to modeling correlation and/or random effects because a simpler analysis that ignores correlation works as well. When the data are unbalanced or missing, the linear mixed model approach is preferable. Emphasis is placed on phase I control chart applications.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados