| Summary: | This work is a simulation study to investigate the sensitivity of multivariate control charts for monitoring mean vectors in a bivariate Gaussian process with individual observations. The multivariate cumulative sum (MCUSUM), the multivariate exponentially weighted moving average (MEWMA) and Hotelling’s T charts are selected for analysis due to their common dependency on the noncentrality parameter. The chart performance is evaluated through the average run length (ARL) or the average time to signal. The impact of utilising in-control limits computed from known parameters or Phase I sample estimates is considered for mean vector shifts. Although designed to monitor mean vectors, the sensibility of the control charts is additionally analysed through different variability sources, including the mixing effect of mean vector shifts with increasing variances or positive autocorrelation in the out-of-control process.
|
|---|
