The Future of Process Monitoring

1
The Future of Process Monitoring

• Bill Woodall
• Virginia Tech
• bwoodall_at_vt.edu

2
Woodall, W. H., and Montgomery, D. C. (1999),
Research Issues and Ideas in Statistical Process
Control, Journal of Quality Technology, 31,
376-386. http//filebox.vt.edu/users/bwoodall/Res
earch20Issues20and20Ideas.pdf Stoumbos, Z.
G., Reynolds, M. R., Jr., Ryan, T. P., and
Woodall, W. H. (2000), The State of Statistical
Process Control as We Proceed into the 21st
Century, Journal of the American Statistical
Association, 95, 992-998.
3
Topics

• Profile Monitoring
• Health-related Monitoring and Public Health
  Surveillance
• Other areas Change point analysis, multi-stage
  processes, and the effect of estimation error.

4
Profile MonitoringThe
objective is to monitor functional data over
time. Woodall, W. H., Spitzner, D. J.,
Montgomery, D. C., and Gupta, S. (2004). Using
Control Charts to Monitor Process and Product
Quality Profiles, Journal of Quality Technology
36, 309-320. http//filebox.vt.edu/users/b
woodall/WOODALL20et20al.pdf
5
Linear Profile Data Framework
response (Y)

n10
explanatory variable (X)

j k
j 2
j 1
time
j 1,2,,k sample profiles ngt2 observations in
each profile
6
Example Vertical Board Density Profile Data from
Walker and Wright (JQT, 2002)We have 24
profiles of vertical density, each profile
consists of n 314 measurements.
7
Two Standard Approaches

• Fit a parametric model to each profile and
  monitor each parameter with a separate chart or
  use a multivariate chart based on the vectors of
  parameter estimators.
• Smooth the profile data at each time period and
  control chart various metrics designed to detect
  specified changes in profile shape.

8
Why use wavelets to model profiles?

• Parametric models or nonparametric smoothing
  combined with metrics are far easier to use.
• Wavelets are hard to understand.
• Dimensionality can still be high.
• Wavelet coefficients may be not be interpretable.

9
Healthcare Monitoring Woodall, W. H.
(2006), Use of Control Charts in Health Care
Monitoring and Public Health Surveillance (with
discussion), To appear in Journal of Quality
Technology. http//filebox.vt.edu/use
rs/bwoodall/JQT20Proofs.pdf

• David Spiegelhalter and Olivia Grigg, Cambridge
• James Benneyan, Northeastern University and
  Institute for Healthcare Improvement
• M.A. Mohammed, University Hospital of Birmingham,
  UK
• Stefan Steiner, University of Waterloo
• Howard Burkom, Johns Hopkins
• Jim Lucas, Consultant, and Rochelle Watkins,
  Curtin University of Technology, Australia

10
Healthcare applications involve

• More attribute data, often with 100 inspection
  and rare events
• More time-between-events data
• Risk-adjustment
• Over-dispersion and multiple testing
• Spatial data and prospective detection of
  clusters

11
A very useful referenceLawson, A. B. and
Kleinman, K. (2005). Spatial Syndromic
Surveillance for Public Health, John Wiley
Sons, Inc., New York, NY.
12
Other Areas

• Multi-stage Processes
• Change-point Methods
• Effect of Parameter Estimation
• Jensen, W. A., Jones-Farmer, L. A., Champ,
  C. W., and Woodall, W. H. (2006), Effects of
  Parameter Estimation on Control Chart
  Performance A Literature Review, To appear in
  the Journal of Quality Technology.
• http//filebox.vt.edu/users/bwoodall/Effec
  t20of20Parameter20Estimation.pdf

13
There are many important research and application
opportunities in profile- and health-related
monitoring.