Calvin, Thomas W. (1993, ASQC) IBM Corporation; Poughkeepsie, NY
Statistical process control (SPC) can help the understanding and implementation of total quality management and continuous improvement. However, traditional SPC tools may be inappropriate when extra variation in a process is not removed, perhaps because the process is new or the causes of variation have not yet been isolated. When the between- subgroup variation is large and random, distribution-free procedures may be useful. If the data are stable, a histogram determined from a stem and leaf plot will indicate whether the data are symmetrical or skewed. For symmetrical data, the control limit is expressed in terms of the median and the interquartile range. For samples that are skewed, the Rosenbaum Count Test assesses the significance of the skew. An example uses a database of 200 order filling times. The data are highly skewed, as demonstrated by a median of 31 with upper and lower quartile values of 50.5 and 22. The resulting sample upper control limit of 366 and the true population upper limit of 413 show that highly skewed data are too erratic for the usual SPC recommendation of 25 data points. Distribution-free techniques are alternatives to the indiscriminate use of normal limits.
Continuous improvement (CI),Control limits,Statistical process control (SPC),Statistics