scientific article; zbMATH DE number 1107639
The \((\overline{x},s)\) control chart is a new type of control chart in which the states of the process are represented by a succession of points \((\overline{x},s)\) on a rectangular coordinate graph, where \(\overline{x}\) is the sample mean and \(s\) is the sample standard deviation with divisor \(n\). By using the \((\overline{x},s)\) control chart, a significant departure of the process from its null state may easily be seen. The main problem is how to specify the control limit of the \((\overline{x},s)\) control chart to give a specified probability of Type I error. In this paper, the difference between the null distribution and the estimated distribution in its present state is evaluated by means of Kullback-Leibler information. The asymptotic distribution of the Kullback-Leibler information statistic for the normal distribution is the chi-squared distribution with two degrees of freedom. However, the sample size used in the control chart is usually small, in which case the chi-squared distribution with two degrees of freedom is not appropriate. Instead we derive an alternative chi-squared approximation to the distribution of the Kullback-Leibler information statistics. We then use this distribution to develop a procedure for determining the control limit of the \((\overline{x},s)\) control chart taking into account the required probability of Type I error. The multiple decisions involved in detecting changes in the process are also discussed. For the entire collection see [Zbl 0867.00026].
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4373576)
