B1852
Title: Multivariate statistical quality control for autocorrelated data
Authors: Stefanos Voutsinas - Athens University of Economics and Business (Greece)
Ioulia Papageorgiou - Athens University of Economics and Business (Greece) [presenting]
Abstract: With new technologies introduced in monitoring industrial processes, such as automatic sensor technology and sophisticated software, the data collected to be used for statistical analysis and control of a process, are most often multivariate today. Moreover, the time between observations can be very short. These two factors result in multivariate autocorrelated data. Standard methods for statistical quality control can be problematic or inferior with respect to their performance in such applications, and false conclusions may be drawn. One of the leading approaches to treat the existence of autocorrelation in the univariate case is the model-based, where a time series model is fitted to the data first, and all standard techniques are then implemented to the residuals instead of initial data. Attempts to extend these methodologies to the multivariate case have been reported to present both difficulties in practical use and questionable results in efficiency. We present two model-free approaches proposed to cope with autocorrelation and assist in monitoring a process.
