Hart Ph.D., Marilyn; Hart Ph.D., Robert; Philip Ph.D., George (1991, ASQC) University of Wisconsin, Oshkosh, WI
This abstract is an edited version of the authors' original.
A control chart is a statistical tool used to aid in keeping the product uniform. The objective of this paper is to present simple but more powerful tests to detect the lack of within-subgroup homogeneity so the control chart can signal changes in the process and corrective action can be taken as needed to make the product statistically uniform.When different "populations" are mixed within a sample, artificially inflated control limits will result, rendering the control chart ineffective. It is imperative to keep the product within subgroups as homogeneous as possible. The most commonly used test for lack of within-subgroup homogeneity is "15 or more consecutive points within � 1 sigma limits of the mean". However, this criterion often fails to detect the lack of within-subgroup homogeneity.The goal of the authors' research is to develop tests for the control chart to detect this mixing of populations within a sample. Data from the literature and simulated data that lacks within-subgroup homogeneity to varying degrees are used to compare the relative power of these proposed tests.