Professor Ronald D. ArmstrongDepartment of Management Science and Information SystemsRutgers UniversityJanice H. Levin Building94 Rockafeller RoadPiscataway, NJ 08854-8054USADear Professor Armstrong:I greatly appreciate the opportunity to strongly recommend Dr. Zachary G. Stoumbos forpromotion to Full Professor at Rutgers University. I assumed that he was already a Full Professorbased on his research recognition and visibility in the academic and professional community. Hispromotion is clearly overdue.Dr. Stoumbos’ work has been primarily in applied and industrial statistics, with a major focus ofhis work being in statistical process control (SPC) and sequential analysis. The numerous areasof application include manufacturing, medicine, management, finance, and law enforcement,among others. His recent work has expanded into innovative research in healthcare, classificationanalysis, and mathematical finance. He has extensively published in top-tier journals such as theJournal of the American Statistical Association, Technometrics, Journal of Quality Technology,IIE Transactions, and the two series of Nonlinear Analysis.The contributions of Dr. Stoumbos to SPC are major, both from the technical as well as practicalperspective. They place him at the forefront of his field as likely the leading researcher of hisgeneration. His papers tackle very important issues in adaptive and dynamic sampling, serialcorrelation, robustness to nonnormality, nonparametric methods, multivariate and multiparameterSPC, and process capability analysis. His investigations are founded on analytical results andhighly accurate numerical methods based on Markov chain modeling and stochastic integralequations, with the occasional use of empirical studies when necessary. I will comment onseveral areas of his work.Traditional process monitoring methods are based on fixed samples taken at regular pre-specifiedpoints in time. Adaptive methods employ variable sampling rate policies in the application ofcontrol charts in order to achieve much better response rates with both standard Shewhart chartsand more sophisticated CUSUM and EWMA charts (see papers #7 and #20). Modern dataacquisition technology, now used in many applications, permits further improvements in theperformance achieved by the SPRT and GSPRT charts developed by Dr. Stoumbos — from firstprinciples — using decision methods based on dynamic sampling. These control charts achievethe fastest and most reliable response rates to date. Papers #32 and #33 are considered seminalon the topic. They have been widely cited and have spurred significant additional research bymany other engineers and statisticians on this important subject, as well as numerous industrialand medical applications of this methodology. This can be seen by a quick search of the ScienceCitation Reports. Dr. Stoumbos and his coauthors have brought adaptive and dynamic controlchart methods from the realm of abstract theory into that of practical SPC tools that areincreasingly being used by major companies, such as Chevron, Exxon-Mobile, Monsanto, andGeneral Electric, and healthcare organizations and providers, such as the Centers for DiseaseControl and Prevention, Mayo Clinic, and Massachusetts’s General Hospital.Multivariate methods have been a major deficiency in quality improvement. Even though manyproducts have multiple quality characteristics and are best studied by multivariate tools, and eventhough multiple processes can be controlled more effectively as a whole rather than as a series ofparallel uncorrelated processes, practice tends to use Dr. Shewhart’s 1931-vintage control charts.Part of the problem is inescapable; there is no multivariate equivalent to the optimality of aunivariate CUSUM chart, but part is due to the paucity of dependable and practical multivariatemethods. Dr. Stoumbos’ collection of parametric and nonparametric papers on detection anddiagnosis in the multivariate arena (see #1, #13, #17, #24, and #29) makes a substantialcontribution to address this major lack and even sets the record straight on some questionablepopular “academic” methods proposed in the last ten years, such as “simplicial data depth.” Hismultivariate papers go a long way towards connecting with the world of practitioners, who arenow realizing the value of this work. With the advent of the 21st century, numerous companiessuch as General Electric, Miele, and Motorola, and agencies such as NASA, the FAA, and IATAhave embraced multivariate SPC and are spearheading its use.One of the fundamental dogmas of SPC is the concept of rational subgroups. Dr. Stoumbos’recent work in paper #9 (with Dr. Reynolds) radically annuls this dogma and renders much of theconventional SPC theory and practices misguided. This seminal work overturns widely adoptedpractices that go all the way back to the early 1930s and Dr. Shewhart himself — the father ofSPC. Thus, it is gratifying that this paper won the 2004 Brumbaugh Award. This prestigiousaward has been given annually since 1949 by the American Society for Quality (ASQ) for thepaper making the greatest single contribution to the development and application of qualityimprovement. It is the leading award of ASQ. It is impressive that Dr. Stoumbos was presentedwith the Brumbaugh Award in the presence of over two thousand attendees, alongsidedistinguished awardees such as Dr. S. Toyoda (Chairman, CEO, and Founder of Toyota MotorCorporation), Dr. F. W. Breyfogle III (President and CEO of Smarter Solutions, Inc.), and Dr. Y.Kondo (Professor at MIT, former Dean of Engineering at Kyoto University, Japan, and formerPresident of the Japanese Society for Quality Control (JSQC)), and the 2005 ASQ ShewhartMedalist Dr. D. M. Hawkins (Distinguished Professor at University of Minnesota).Some of the strongest technical work of Dr. Stoumbos can be found in a series of importantpapers on monitoring discrete process variables that can be modeled with Bernoulli and binomialdistributions. The discreteness and skewness of these variables makes traditional process controlapproaches ineffective. In these papers, exact expressions were derived and evaluated for theproperties of the Bernoulli CUSUM, Binomial CUSUM, and SPRT charts for a proportion