The use of the shifting mean heuristic in statistical quality control

Abstract

Demonstrates the application of the recently proposed shifting mean heuristic to statistical quality control situations. Proposes the heuristic search procedure primarily as a tool for retrospective (or post mortem) data analysis, retrospectively examining a known data set to establish shifts of process mean, as illustrated in the traditional Manhattan diagram (i.e. a plot of observations with superimposed sub‐process means). However, because the procedure can be operated automatically, it is suggested that where sampling rates are relatively slow it could also be used for online quality monitoring by providing a dynamic Manhattan diagram which changes shape when a newly developed series of observations establishes a shift from a pre‐specified reference level or target value. Briefly reviews other approaches to establishing shifts in process mean, describes the concepts underlying the heuristic search procedure, and reviews the statistical considerations involved. Compares the procedure with other approaches, and provides examples of the heuristic search procedure operating as a tool for retrospective data analysis using three time series kindly provided by the editor which were analysed “blind” by the author. The editor’s comments on the results are appended.