Advisor

Schaffer, Jay R

Committee Member

Mundfrom, Daniel J

Committee Member

Perett, Jamis J

Committee Member

Heiny, Robert L

Department

Applied Statistics & Research Methods

Institution

University of Northern Colorado

Type of Resources

Text

Place of Publication

Greeley (Colo.)

Publisher

University of Northern Colorado

Date Created

12-1-2009

Genre

Thesis

Extent

133 pages

Digital Origin

Born digital

Description

In statistical process control (SPC), continued development of techniques look for new monitoring charts for processes with multiple correlated variables. Two such charts are the multivariate exponentially weighted moving standard deviation (MEWMS) and multivariate exponentially weighted moving variance (MEWMV). Originally developed by Huwang, Yeh, and Wu (2007), and furthered by Hawkins and Maboudou-Tchao (2008), these control charts monitor the trace elements of the respective covariance matrices for a change in values of the multivariate process using individual observations. Originally, control chart parameters were developed during the simulation process for p = 2 and p = 3 process variables. Using computer simulations of 10,000 replications, further development of the MEWMS and MEWMV used p = 5 and p = 10 correlated variables with individual observations to develop control chart parameters and determine the sensitivity of the MEWMS and MEWMV compared to the multivariate CUSUM and MEWMA charts in their detection of a singular shift of mean, a singular change in variance or a combination of the two. Initial findings from this dissertation suggest that both the MEWMS and MEWMV control charts are highly sensitive to small changes of a single element in the covariance matrix and sensitive to changes in a single element of the observation vector. When comparing the MEWMS and MEWMV control charts to the MCUSUM and MEWMA control charts popularly used today, it was found that the MEWMS and MEWMV control charts are less sensitive to mean shifts than the MCUSUM or MEWMA. However, it was also shown that the MEWMS and MEWMV are much more sensitive to changes in elements of the covariance matrix than the MCUSUM, while the MEWMA control chart is insensitive to any changes of variance elements.

Notes

APPENDICES2.zip other Appendices B&C

Degree type

PhD

Degree Name

Doctoral

Language

English

Local Identifiers

Eshelman_unco_0161N_10017

Rights Statement

Copyright is held by author.

Share

COinS