A comparison of phase I control charts

Abstract

A carefully done Phase I analysis is a vital part of an overall statistical process control and monitoring regime. Distribution-free control charts can play a useful role in this analysis, as a parametric model assumption often cannot be adequately verified. The performance of two distribution-free charts for the location – the mean-rank and the median chart – are compared in this paper. For benchmarking reasons, the parametric -chart is included in the comparisons. It is seen that the distribution-free charts are in-control robust, whereas the -chart is not, and they both perform similarly to the -chart for normally distributed data. However, for non-normal data, they both outperform the -chart. The results provide evidence in favour of using distribution-free Phase I charts in practice. Concluding remarks and suggestions for future research are given.

Keywords

• retrospective phase
• nonparametric
• simulation