William G. Hunter Award 2010: Tom Nolan

The recipient of the 2010 William G. Hunter Award is Tom Nolan. The Statistics Division of the American Society for Quality (ASQ) established the Hunter Award in 1987 in memory of the Division's founding chair to promote, encourage and acknowledge outstanding accomplishments during a career in the broad field of applied statistics. The attributes that characterize Bill Hunter's career - consultant, educator for practitioners, communicator, and integrator of statistical thinking into other disciplines - are used to help decide the recipient.

Acceptance speech by Tom Nolan, 2010

Thanks to the selection committee for choosing me for this award in honor of William Hunter and to Davis Balestracci for taking the time to put together the materials for my nomination. I especially want to thank a statistician whom I have never met and whose name I do not know. Let me explain. In 1972 I was finishing a master's degree in statistics and looking for a job. I saw an advertisement for a position in the statistics group at the Agriculture Marketing Service in the US Department of Agriculture. I applied for the job, got an interview, and shortly after received a letter that someone else had been chosen and had accepted the position. Oh well. About two weeks later I received a phone call asking if I was still interested in the position because the person they hired had changed his mind about taking the job. I gladly accepted and became part of the statistics group of AMS.

The group was headed by Richard Bartlett a graduate of Virginia Tech. Dick had established a culture in the group that aligned well with the culture that Bill Hunter promoted at Wisconsin. We were a service organization whose mission was to use applied statistics to make the program divisions more effective. Dick also promoted a culture of learning and collaboration. Being part of the group introduced me to the world of statisticians working in government and industry who embodied Bill Hunter’s attributes, people such as Brian Joiner, Gerry Hahn, and Ron Snee who have received the Hunter Award. This was a very good start. I was lucky. Two members of that group Lloyd Provost and Ron Moen are presently colleagues of mine in API, a 38 year relationship that began at USDA because someone turned down the job that I filled.

Since then I have worked in many different industries applying statistical methods to the improvement of quality and productivity. To be effective in a variety of settings I have relied on three building blocks for statistical thinking that I offer to you for your consideration.

1. The first is W. Edwards Deming’s distinction between enumerative and analytic studies. An enumerative study is one in which action will be taken on the frame, for example valuing an inventory or certifying a batch of material. An analytic study is one in which action will be taken on a system to improve its performance in the future. As Deming often said the aim of an analytic study is prediction. Will the change continue to be an improvement in the future? The degree of belief in these predictions - high, medium, low - cannot be quantified using statistical methods alone. The subject matter expert determines whether the conditions of the tests are varied enough to make the results useful for decision-making about the future. Developing and testing new products is clearly in the realm of analytic studies.
2. The second building block is Walter Shewhart's approach to classifying sources of variation into common causes and special causes. In my university education we started with the assumption of a distribution underlying the data we were collecting. Implicitly we were assuming that the data came from a stable system. In my practice which consists of working primarily on analytic studies, I realize that the data are coming from a system that may be unstable and for which no stable distribution yet exists. Shewhart developed the control chart method to sort out whether a system was stable or unstable. He sought to minimize the economic loss of overreacting or under reacting to variation. Such unnecessary losses occur even today for example during the review of the results of medical diagnostic tests or the review of quarterly results in financial measures.
3. The third building block is Ginichi Taguchi's distinction between control factors and noise factors. This distinction gives us an effective way to use experimental design to reduce common cause variation in systems. In the field of education right now there is a theme that good teaching is the key to student achievement because of research that shows a correlation between teaching competency and student achievement. One reaction is to reward good teachers and fire bad ones. Another perspective following Taguchi's line of thinking is to view variation in teacher competency as a noise factor – teachers will always vary in their competency - and then manipulate control factors such as curriculum design, new teacher development, and parent support to mitigate the effect of variation in teacher competency on student performance.

There will always be tremendous opportunities for statisticians who are willing to follow the lead of William Hunter to be service oriented and be willing to collaborate effectively with other disciplines to solve the complex problems of the day.

Finally one request, if anyone here turned down a job in the statistics group at the US Department of Agriculture in 1972, please come forward and introduce yourself. I would like to thank you in person.