Three suggestions for improving control charting procedures

Discusses some of the practical difficulties commonly experienced with Shewhart control charts, and makes three suggestions for overcoming these difficulties. The first suggestion is to distinguish the “out‐of‐control” points by some means other than control lines – perhaps by ringing them on the graph – and to avoid the phrases “in control” and “out of control”. Argues that this would avoid some common misconceptions. The second suggestion is a procedure, method A, for calculating action and control levels based on the median and the quartiles of the points plotted. This is a much simpler and more general procedure than the usual ones. Argues that method A has a number of advantages over the standard procedures (for mean charts, range charts, P charts, C charts and so on) provided that a long sequence of sample data is available. The third suggestion, method B, is an alternative to the second: this is a bootstrapping procedure which is conceptually simple but requires considerable computer power for implementation. Method B also has advantages over the standard procedures whenever the points plotted are calculated from a random sample of data. In addition, unlike method A, it is feasible if there is only a short sequence of samples of data available. All three suggestions are designed to simplify the idea of control charts, and so are likely to facilitate the learning process.