Back to Conference information.
The talk will review some aspects in the history of the application of statistical methods in industry during the previous century. The focus will be on statistics understood as a problem-orientated discipline providing concepts and methods to be used as tools for describing and analyzing variation and the transfer of variation, in order to transform data to relevant information.
Generic statistical concepts and techniques are most often a result of problem-driven activities in a specific application situation. Throughout history, statistical methods developed in one field proved to be useful in another — although the transition sometimes has been rather slow. There have been major interactions with social sciences, genetics, physics, chemistry, biology and the agricultural field leading to valuable statistical concepts and tools. Also, tools and techniques in the field of industrial statistics originate from the work of skilled statisticians in industry, using their core competences in a multidisciplinary team combining specialist knowledge and insight from relevant fields. The activities in the middle of last century shows many examples of such an interaction where the ability of the statistician to analyze and understand the nature of the sources of variation present in the problem at hand and their relative importance did result in what is now considered a generic statistical tool.
Also today, statistically based procedures are developed in industry aiming at solving a problem at hand. As a consequence of the rather slow transfer of knowledge and insight in the variety of statistical concepts and tools, such activities sometimes result in ``reinvention of the wheel''. Even within the field of industrial statistics, transfer of knowledge between subject areas might be better. Thus, it is noteworthy that although the concept of ``blocks'' has long been appreciated and accounted for when designing and analyzing industrial experiments, most SPC applications still do not consider the possibility of between subgroup variability originating from ``common causes''.
The inventory of the statistical toolbox today contains a rich selection of tools of various levels of sophistication that are tailored for a multitude of different manifestations of variation. The abundance of data, and possibilities of processing the data in current business and industrial environment presents a challenge for the statistical profession to provide value adding services at operational, tactical, and strategic levels. However, to respond to this challenge, I believe that it is imperative that we use and enhance our core competences in understanding variation and use our communication and listening skills in cooperating with the specialists in the field. That we cooperate in prioritizing needs, and keep the problem in focus when selecting an appropriate tool from the toolbox, rather than letting our favourite tool, or the currently popular buzz-word have higher priority than the customer's problem.
Poul Thyregod (http://www.imm.dtu.dk/~pt) was educated in mathematical statistics at University of Copenhagen (M.Sc. 1966, Ph.d 1970).
1970-71 Visiting at School of Statistics, University of Minnesota.
Since 1971 employed at the Technical University of Denmark, Institute of Mathematical Statistics and OR (IMSOR), now Informatics and Mathematical Modelling (IMM).
Cooperated with various industrial sectors and governmental agencies in a number of projects involving use of statistical methods. Active participant in International Standardization activities, chairman of Strategic Policy Advisory Group (1994-1999).
Member of the editorial board of Quality and Reliability Engineering International. Board member of Danish Society for Applied Statistics (FAST) - which is also the danish branch of ENBIS-DK.
Authored publications on topics in acceptance sampling and quality monitoring, analysis of reliability data, application of statistical methods in a wide range of application fields.
AbstractIn this talk we shall describe how business consulting and statistics can interact in a synergic way. We shall first describe how a business consulting group, such as the Information Risk Management group of Kpmg led by me, develops and operates a consulting activity, partnerships and strategies. In particular, we shall emphasize the growing role of Information Technology driven consulting areas.
We then dwell on the main areas of consulting, at the moment under skill shortage, for which we claim a future growth. Again a particular attention will be dedicated to information technology related areas.
We shall then turn our attention to the perception of statistics and data analysis in daily consulting activities, both from the client's and the consultants' side. This will give us an opportunity to support our clients to understand and to develop information systems, integrated with their processes, in the areas where statistics and data analysis can have an important scope and range of application.
Finally, we shall describe, by means of a real example, how a consulting company and an academic department of statistics can cooperate towards the common goal of letting the awareness of the importance of statistics to grow in today's society.
The example concerns a comparative study of data mining softwares. The talk will not enter technical details of the study but, rather, will describe how opportunities have emerged from this cooperation for both partners involved, namely, Kpmg and the University of Pavia.
This project is indeed a strategic phase of a global project designed to support clients to build an integrated Business Intelligence solution, using Data Mining techniques as well as Datawarehousing ones.
Davide Grassano has a degree in economics with Genoa University, he is qualified labour consultant and Italian accountant. Currently, he is head of Information Risk Management for Italy. He is experienced in security services and IS audit in financial sector and ICE. Furthermore, in project control and risk management for client operating in different business sectors; most IRM services such as system integration control, privacy, security, pki and Business continuity services.
This talk will focus on two examples of the use of simple statistical methods in the pharmaceutical manufacturing industry. Both relate to making measurements, the importance of which is commonly underestimated, in the author's experience.
The first example is concerned with the use of Shewhart control charts in an analytical chemistry QC laboratory. The nature of the chance variation that affects measurements carries implications for:
- how control limits are drawn
- how many replicates are made
- how these are measured.
The implications for the effectiveness of the charts in monitoring the analytical system and for laboratory workload will be illustrated and discussed.
The second example involves a simple comparative study of the results produced by two QC laboratories. The way the study is carried out carries implications for the results of the paired t-test that is used to test for a relative bias between the laboratories. It will be shown that the Ã¢â‚¬ËœnaturalÃ¢â‚¬â„¢ way to conduct the study virtually guarantees the detection of a relative bias when none exists and that the more data that are collected the more likely this is to occur.
In both cases the key issue is the nested nature of the chance variation affecting the system under study. In both cases the application of the methods in the "natural" way leads to erroneous decisions being made, with potentially serious practical consequences. The moral of the two stories is that the application of even simple statistical methods is rarely trivial and that we must emphasize statistical thinking in teaching engineers and scientists. Statistical thinking means focusing on the nature of the chance variation that influences the system under study.
The two examples outlined above are based on the author's industrial consultancy experience. The final part of the talk will use the examples as a basis for reflecting on that experience and will try to draw out some lessons about statistical consultancy. Much of what I will say is implicit in the title of the talk.
Eamonn Mullins is a Senior Lecturer in Statistics at the Department of Statistics, Trinity College, Dublin, Ireland. Since 1999 he has also been Head of the Department. He holds a B.Sc. degree in Experimental Physics and Mathematics from University College, Dublin and an M.Sc. in Applied Statistics from Trinity College. He has lectured there since 1975. His principal interests are in the application of statistical methods in industry and he has wide experience in industrial consultancy, particularly in short course delivery. He was a member of the Irish National Accreditation Board (NAB) from 1996 to 2001 and is currently the Chairman of its Appeals Committee. NAB is the national body responsible for accreditation in accordance with the EN 45000 series of European standards and the relevant ISO standards and guides. His book "Statistics for the Quality Control Chemistry Laboratory" is to be published by the Royal Society of Chemistry in Spring/Summer 2003.