A0777
Title: Comparing the treatment of old data in some self-starting control charts
Authors: Claudio Giovanni Borroni - University of Milano - Bicocca (Italy)
Manuela Cazzaro - University of Milano-Bicocca (Italy) [presenting]
Paola Maddalena Chiodini - University of Milano-Bicocca (Italy)
Abstract: Self-starting control charts do not distinguish between Phase I and Phase II in the statistical control task. Nonetheless, control cannot really start unless a minimum number of readings is collected and, according to what is reported by many authors, such minimum needs often to be raised to get a suitable performance. That depends mostly on how the chart tentatively separates between the new and the old observations and on how it is instructed to learn from the second kind of data. Being that, in some applications, the availably of a large reference sample (i.e. collected under normal conditions) is often prevented, the aim is to compare the strengths of some recent proposals when the control needs to be started after very few initial readings are gathered. The analysis is based on some simulations from distributions with different shapes; in addition, it considers different kinds of shifts from the in-control to the out-of-control situation and different instants when that shift occurs after control has started.
