A nonparametric exponentially weighted moving average control scheme

Abstract

A common assumption when evaluating the properties of Exponentially Weighted Moving Average (EWMA) control procedures for controlling the process mean is the observations are normal with known variance. In this article we propose a nonparametric control procedure that can be used when the underlying distribution is not know there is not enough information on the variance or shape of the distribution. Its average run length properties are less affected than the corresponding parametric EWMA procedure when autocorrelation between the observations is present. The procedure nonparametric EWMA procedure is based on Wilcoxon signed-rank statistics ranking is within groups. Our simulation results show that the proposed control procedure is less efficient than the parametric -EWMA procedure when the distribution is not and it can be considerably more efficient than the parametric procedure for heavy- distributions. The proposed procedure is insensitive to misspecification of the van and its ARL properties of the control procedure are affected relatively little by cho the weighting parameter λ.

Keywords:

• Control Charts
• Average Run Length
• Nonparametric Statistics
• Wilcoxon S Rank Statistic
• Process Control