White, Dr. Edna M.; Schroeder, Dr. Roger G. (1988, ASQC) Northeastern University, Boston, MA; University of Minnesota, Minneapolis, MN
This paper compares the performance of resistant and non-resistant measures in process control of variables. The effectiveness of the most commonly used measures, the non-resistant mean and range, are compared to two recently suggested measures, the resistant median and Q-Spread. Comparison is made under various simulated situations involving seven underlying distributions and two process conditions. The simulations studies show a tradeoff between the sensitivity of the resistant measures and the robustness of the non-resistant measures. Evaluations of these tradeoffs in the context of process control leads to a set of decision rules to aid in the selection of a control chart technique.
Chemical and process industries,Process capability studies