The AAPS Journal

, Volume 7, Issue 1, pp E106–E117 | Cite as

Effect of multivariate process instability on principal component analysis: A case study

  • Donald S. Holmes
  • A. Erhan Mergen


With the rising use of principal component analysis/partial least squares (PCA/PLS) in the process analytical technology (PAT) initiative of the pharmaceutical industry, it seems appropriate to view that approach from a statistical process control (SPC) perspective. The purpose of this study was to demonstrate the effect of process instability (ie, state of statistical out-of-control) on use of PCA/PLS. The demonstrated differences in results should encourage PCA/PLS users to incorporate SPC as an active part of their process analytical control (PAC) toolkit to check for stability prior to drawing conclusions based on PCA/PLS analysis.


principal component analysis partial least squares covariance matrix mean square successive differences (MSSD) 


  1. 1.
    Johnson R, Wichern D. Principal Components. In:Applied Multivariate Statistical Analysis. 3rd ed. Upper Saddle River, NJ: Prentice Hall; 1992:356–357.Google Scholar
  2. 2.
    Holmes D, Mergen E. A multivariate test for randomness.Qual Eng. 1998;10:505–508.CrossRefGoogle Scholar
  3. 3.
    Hald A. Statistical Control. In:Statistical Theory with Engineering Applications. New York, NY: John Wiley & Sons; 1952:357–360.Google Scholar
  4. 4.
    Holmes D, Mergen E. Improving the performance of T-square control chart.Qual Eng. 1993;5:619–625.CrossRefGoogle Scholar
  5. 5.
    Kramer C, Jensen D. Fundamentals of multivariate analysis. II. Inference about two treatments.J Qual Technol. 1969;1:189–204.Google Scholar
  6. 6.
    Wilks S. Certain generalizations in the analysis of variance.Biometrika. 1932;24:471–494.Google Scholar

Copyright information

© American Association of Pharmaceutical Scientists 2005

Authors and Affiliations

  1. 1.Stochos IncSchenectady
  2. 2.Rochester Institute of TechnologyRochester

Personalised recommendations