Abstract:
|
The knowledge of the underlying parameters, or the existence of a sufficiently large reference sample from which they can be estimated, is required for implementing most control charts. Estimation of parameters is known to degrade charts performance. This paper studies the effects of estimation of parameters on the performance of the nonparametric multivariate sign EWMA (MSEWMA) chart by Zou and Tsung (2011). First, the effects of using their known parameter control limits are studied in case when the parameters are estimated from a relatively small Phase I (reference) sample. It is seen that in this case the chart performance is highly degraded, in that, many more false alarms are observed than what is nominally expected. Next, using simulations, correct control limits are obtained that achieve a desired in control ARL when parameters are estimated from a given size reference sample. The out-of-control performance of the MSEWMA chart with the corrected control limits is also briefly studied. The use of the proposed corrected control limits is recommended when the reference sample is not too large.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.