A performance analysis of multivariate nonparametric control charts

Pike, Levi (2017) A performance analysis of multivariate nonparametric control charts. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (2MB)

Abstract

Robust and efficient multivariate control charts are not common in literature. This report explores the versatility of the few distribution-free, nonparametric multivariate Statistical Process Control (MSPC) charts suitable for average run length (ARL) analysis. Current datasets are becoming increasingly complex, large, and less likely to follow distributional properties required for traditional parametric statistics, a fact especially true for a multivariate setting. The purpose of our study is to compare the newest available methods, not previously compared with one another in cases and data structures not yet explored. Due to the versatility and robustness of the types of data these methods can accommodate, finding real world applications is trivial. The five methods applied here are able to exploit different types of changes to the structure of a distribution, rather than simply detect a mean shift. These methods have similar features, able to avoid lengthy data-gathering steps, and applicable in short-run and start up situations. By establishing cut-off values simultaneously based on input observations, rather than beforehand, the methods are applying data-dependent control limits which shows their truly distribution-free property. Some of the current areas of improvement continue to be on creating more computationally efficient algorithms for these methods.

Actions (login required)

View Item