Robust Analysis of Variance: Process Design and Quality Improvement

We discuss the use of robust Analysis Of Variance (ANOVA) techniques as applied to quality engineering. ANOVA is the cornerstone for uncovering the effects of design factors on performance. Our goal is to utilise methodologies that yield similar results to standard methods when the underlying assumptions are satisfied, but are also relatively unaffected by outliers (observations that are inconsistent with the general pattern in the data). We do this by utilising statistical software to implement robust ANOVA methods, which are no more difficult to perform than ordinary ANOVA. We study several examples to illustrate how using standard techniques can lead to misleading inferences about the process being examined, which are avoided when using a robust analysis. We further demonstrate that assessments of the importance of factors for quality design can be seriously compromised when utilising standard methods as opposed to robust methods.

[1]  W. Sharpe.  Mean-Absolute-Deviation Characteristic Lines for Securities and Portfolios , 1971 .

[2]  Manfred W. Padberg,et al.  The Finite Sample Breakdown Point of #8467;1-Regression , 2004, SIAM J. Optim..

[3]  Neil R. Ullman,et al.  Signal-to-noise ratios, performance criteria, and transformations , 1988 .

[4]  R. A. Rigby,et al.  A semi-parametric additive model for variance heterogeneity , 1996, Stat. Comput..

[5]  Frederick R. Forst,et al.  On robust estimation of the location parameter , 1980 .

[6]  Vic Barnett,et al.  Outliers in Statistical Data , 1980 .

[7]  R. Carroll.  Robust Methods for Factorial Experiments with Outliers. , 1979 .

[8]  Genichii Taguchi,et al.  Introduction to quality engineering. designing quality into products a , 1986 .

[9]  Genichi Taguchi,et al.  Taguchi methods : design of experiments , 1993 .

[10]  R. Koenker,et al.  The Gaussian hare and the Laplacian tortoise: computability of squared-error versus absolute-error estimators , 1997 .

[11]  D. Lax.  Robust Estimators of Scale: Finite-Sample Performance in Long-Tailed Symmetric Distributions , 1985 .

[12]  Gordon K. Smyth,et al.  Generalized linear models with varying dispersion , 1989 .

[13]  A. Gelman.  Analysis of variance: Why it is more important than ever? , 2005, math/0504499.

[14]  Raghu N. Kacker,et al.  Performance measures independent of adjustment , 1987 .

[15]  J. Simonoff.  Outlier detection and robust estimation of scale , 1987 .

[16]  D. Ruppert,et al.  [Signal-to-Noise Ratios, Performance Criteria, and Transformations]: Discussion , 1988 .

[17]  P. L. Goldsmith,et al.  Critical Analysis of Factorial Experiments and Orthogonal Fractions , 1973 .

[18]  Stephen F. Filippone.  Using Taguchi methods to apply the axioms of design , 1989 .

[19]  Jim Freeman,et al.  Outliers in Statistical Data (3rd edition) , 1995 .

[20]  Jeffrey S. Simonoff,et al.  A comparison of robust methods and detection of outliers techniques when estimating a location parameter , 1984 .

[21]  Zachary G. Stoumbos,et al.  Robustness to Non-Normality of the Multivariate EWMA Control Chart , 2002 .

[22]  Soren Bisgaard.  Quality Engineering and Taguchi Methods: A Perspective , 1990 .

[23]  David M. Rocke.  Xq and Rq charts: Robust control charts , 1992 .

[24]  Joseph J. Pignatiello,et al.  TOP TEN TRIUMPHS AND TRAGEDIES OF GENICHI TAGUCHI , 1991 .

[25]  P. J. Huber.  Robust Regression: Asymptotics, Conjectures and Monte Carlo , 1973 .