Home » Featured

The World Is Calling; Should We Answer?

1 May 2012 3,950 views 5 Comments
Editor’s Note: This article was adapted from the Deming Lecture, given at the 2011 Joint Statistical Meetings in Miami Beach, Florida.

    Roger Hoerl speaks during JSM 2011 in Miami Beach, Florida.

    A watershed event in human history occurred in Copenhagen, Denmark, from May 24–28, 2004, during a meeting referred to as the Copenhagen Consensus. Bjorn Lomborg, the Swedish economist, brought together a “blue ribbon” panel, including three Nobel laureates, from around the globe to discuss the world’s greatest challenges. Rather than prioritizing problems, the panel was asked to prioritize potential solutions. Specifically, the panel was asked to answer the following question: If one had $50 billion, how could this money best be spent to improve the human condition? Therefore, the panel had to consider not only how severe the problems were, but also what impact could be made with this level of expenditure. For example, virtually everyone is appalled by human trafficking, but what impact could be made to end it with this amount of investment?

    The panel began with a list of the greatest challenges facing the world, reproduced in Table 1. They then reviewed suggested solutions that had been proposed by globally recognized authorities in each field, which included some estimate of the expected impact of their proposals. Finally, members of the panel met to prioritize the solutions, based on the impact they felt each was likely to have. In his summary of the gathering, Lomborg wrote the following:

    Combating HIV/AIDS should be at the top of the world’s priority list. That is the recommendation from the Copenhagen Consensus 2004 expert panel of world-leading economists.… About 28 million cases could be prevented by 2010. The cost would be $27 billion, with benefits almost 40 times as high….

    What Has Happened in the Battle Against HIV/AIDS Since 2004?

    Interestingly, Lomborg organized a second Copenhagen Consensus meeting in 2008, at which a proposed solution to global malnutrition was the top priority. One might logically ask if perhaps HIV/AIDS is no longer a major issue, given existence of antiretroviral drugs (ARVs). Sadly, the facts paint a different picture. For example, the World Health Organization announced in 2009 that AIDS-related illnesses are now the leading cause of death globally for women of childbearing age (15–44). In the United States, AIDS is now the leading cause of death for African-American women between 25–34 years of age. Few people seem to have taken notice of these trends, sometimes referred to as the “feminization of AIDS.”

    The cover story of the July 20, 2010, New York Times discussed a randomized trial conducted in Malawi. In this study, the treatment group of teenage girls was given $1–$5 per month if they stayed in school, and their parents were given $4–$10 per month for household expenses. The control group of girls received no money, nor did their parents. Girls were randomly assigned to groups. At the end of 18 months, the data showed that girls in the control group had an HIV infection rate 2.5 times that of the treatment group. This difference was statistically significant using any reasonable test. A jugular question that requires answering is whether this result is acceptable. Is it acceptable that a girl’s chances of acquiring HIV are in large part dependent on access to $10 per month?

    Table 1—Copenhagen Consensus, 2004

    My own experiences investigating the AIDS pandemic are discussed at length elsewhere. In short, I received a six-month sabbatical from GE Global Research as part of my Coolidge Fellowship, which I used to study why AIDS seemed to defy resolution, despite the expenditure of billions of dollars, and what might be done about it. After considerable reading; interacting with AIDS activists and researchers; and spending a month traveling through Uganda, Zambia, and South Africa, my co-author, Presha Neidermeyer, and I came to a number of conclusions, including the following:

        Money is necessary, but not sufficient to resolve this pandemic. It is sometimes difficult for westerners to accept that they cannot “buy” a solution to a major problem.
        Similarly, “made in America” solutions often fail outside a U.S. context. Malawi is not Kansas, and solutions that work in Kansas often simply do not work in Malawi.
        This issue is not about charity, but rather about justice. While western celebrities often run charitable events to raise money for worthwhile causes and visit developing countries for photo shoots in impoverished areas, too often such efforts seem more motivated by what researchers call a “god complex,” rather than a desire for justice. A god complex is the subconscious belief that one is fundamentally superior to those in abject poverty, and because such people are typically wealthy and western, they often come as “saviors” rather than servants to the developing world.

    Presha and I discuss a number of Africans we have met personally who we feel are inspirational heroes, each of whom began significant and successful efforts to address AIDS and/or poverty in their communities with no western help. Of course, western aid can enable such people to help thousands, rather than hundreds, of people, so clearly money does have an important role to play. Two such people are:

      Jolly Nyeko, a woman from Kampala, Uganda, who quit a paying job to counsel adolescent victims of sexual abuse on a volunteer basis. She gradually expanded her efforts and organization to include AIDS testing and treatment, legal referrals, education, and micro-loans. Her nongovernmental organization (NGO), Action for Children, has now affected the lives of roughly 30,000 Ugandans.

      Pastor Titus Sithole of Mamelodi, South Africa—a black township of about 1 million people outside of Pretoria—was publicly tested for HIV in front of his 2,000-member congregation, along with all his leaders, to make the point that AIDS is just a disease, and should not be stigmatized. This act of courage was and is virtually unheard of. Titus has subsequently opened an AIDS clinic within the community, as well as a children’s home for AIDS orphans and a primary school with approximately 450 students.

    Addressing Large, Complex, Unstructured Problems

    Given the types of problems the world faces, such as AIDS and the other challenges listed in Table 1, we might ask why statisticians are not working on these problems. The short answer, of course, is that we are. I certainly want to acknowledge the outstanding work being done by statisticians around the globe who are focusing on finding solutions to cancer, AIDS, climate change, providing safe drinking water in resource-limited areas, and so on. They are doing precisely what I am suggesting. Unfortunately, I don’t think there are enough of them. We need to do more, and I think we can do more.

    The types of problems listed in Table 1 are not typical textbook problems in statistics. For example, in virtually every problem in statistical textbooks, there is a correct answer that can be calculated precisely, without any understanding or knowledge of any other discipline. In other words, the problems are well defined and narrow. The same could be said of many other disciplines. While the reasons for such an approach are understandable—certainly we would not ask a Stat 101 student to solve the global AIDS pandemic—over time it creates an expectation that problems should be well defined, of manageable size, and have a “correct” answer—an optimal solution. Of course, the real world does not work this way. The unfortunate result is that researchers often walk past the critical but messy problems begging for a solution, preferring to find much smaller, less important problems that are well defined and solvable in a short timeframe.

    Science and Engineering

    In January of 2010, Susan Hockfield, president of Massachusetts Institute of Technology (MIT) and member of the GE Board of Directors, gave a talk at the GE Global Research Center in Niskayuna, New York. One of the topics she spoke about was the need for better integration between science and engineering. Loosely quoting from my notes, Hockfield stated:

    Around the dawn of the 20th century, physicists discovered the basic building blocks of the universe, a “parts list,” if you will. Engineers said, “We can build something from this parts list,” and produced the electronics revolution, and, subsequently, the computer revolution. More recently, biologists have discovered and mapped the basic “parts list” of life—the human genome. Engineers have said, “We can build something from this list,” and are producing a revolution in personalized medicine.

    Hockfield’s comments on science and engineering reminded me of a recent article by Xiao-Li Meng. In this article, Meng noted that Harvard had recently added a course, Stat 399: Statistical Problem Solving, to its curriculum. This is a course that, in Meng’s words “… emphasizes deep, broad, and creative statistical thinking, instead of technical problems that correspond to a recognizable textbook chapter.” This caught my attention, since the types of problems in Table 1 certainly do not correspond to a recognizable textbook chapter in any discipline.

    Meng’s article also resulted in significant dialogue with a professional colleague, Ron Snee, in which we discussed how one should structure such a course as Stat 399. There does not seem to be any consensus as to how one should attack large, complex, unstructured problems. What approaches to attacking large problems should be included in the course, and how would one know they are the best methods? By what theory would we answer these questions? It seemed to us that while significant theory exists about individual statistical methods, very little theory or research has been developed on how these methods should best be linked and integrated in order to attack large problems. This line of thinking led us to conclude that the science of statistics needed to be integrated with engineering approaches, per Hockfield’s comments. That is, we think there needs to be a greater emphasis on building something of interest to society from the statistical science “parts list” of tools.

    Table 2—Attributes of Large, Complex Unstructured Problems

    Semantics are important in this discussion, so I will define my terms. A problem need not be as difficult as world hunger to be considered large, complex, and unstructured. Snee and I developed some typical attributes of these problems, which are listed in Table 2. Science can be defined many ways, but typically refers to the use of observation and experimentation to discover, understand, and explain natural phenomena. Published definitions of engineering are typically some variation of the following: Engineering is the study of how to best use scientific and mathematical principles for the benefit of humankind. In other words, engineering does not advance scientific laws—science does—but rather attempts to develop a theory of how these laws can be better used for practical benefit—that is, how to build something of importance from the scientific parts list.

    Statistical Engineering

    Based on these definitions, Snee and I defined the term “statistical engineering” as the study of how to best use statistical concepts, methods, and tools and integrate them with information technology and other relevant sciences to generate improved results. We think it is consistent with dictionary definitions of engineering, discussed above. We have further argued that the statistics profession has, in general, primarily focused on advancing statistical science—the development and application of new methods—while not recognizing that statistical engineering is the “other side of the coin” that could enable statistics to have broader societal impact.

    Relating this to Hockfield’s comments, it occurred to us that while there currently is a lot of emphasis within the profession on advancing the “parts list” of statistical methods, there is considerably less emphasis on “building something of importance to society” from this parts list. I refer, for example, to building overall approaches to problemsolving or process improvement that involve multiple methods. Certainly, there are counter-examples such as the Six Sigma methodology, but there is little theory as to how this should best be done (i.e., what works, what doesn’t, and why).

    Many experienced statisticians have figured this out on their own, through trial and error, and applied statistical engineering without labeling it as such. Unfortunately, this has typically been done on an ad-hoc basis, with little theory or research documented as to how such statistical engineering approaches should be deployed. The natural result is that statisticians, even experienced statisticians, have had to “reinvent the wheel” when faced with new problems having the attributes listed in Table 2. For example, I would argue that virtually no rigorous research or accepted theory exists on how to best integrate multiple statistical tools, or how to integrate statistical and nonstatistical tools.

    Since publishing a definition of statistical engineering and discussing it with statisticians in a number of venues, we have been asked one question more than any other. This question has been asked in a few ways, such as: Is statistical engineering just another term for applied statistics? Isn’t this what statisticians have always done? Is this an attempt to rebrand applied statistics? These questions are certainly reasonable and to be expected, as we are using the term statistical engineering in a new way and there is a clear overlap with applied statistics. Using my own career as an example, I think I did a good job of applying the statistical methods I learned in graduate school while an intern with the DuPont Applied Statistics Group in the summers of 1981 and 1982. However, I certainly did not build anything novel from the statistical science parts list of tools; in other words, I applied individual tools to well-defined problems, but did not integrate the tools in any innovative ways to achieve breakthrough results. I did good applied statistics, but no statistical engineering.

    A number of commentators have pointed out that while most academic science departments are becoming narrower and more compartmentalized, the truly groundbreaking research being conducted today is almost exclusively cross-disciplinary in nature. Addressing the large, complex, unstructured problems we currently face will require both breakthroughs in various sciences, including statistics—such as the development of newer and better statistical tools—and also the effective integration of diverse scientific methods into broader engineering approaches. The IBM computer Watson, which successfully competed on the U.S. game show Jeopardy! and is now being applied to problems in public health and consumer finance, is one example. Watson integrates technology from a variety of disciplines, including computational science, natural language processing, statistics, and machine learning, to name a few.

    In summary, it is clear that the world has numerous complex challenges and is in need of leadership to address them. I believe statisticians can answer the call and have broader societal impact, if we choose to. A balanced emphasis on both statistical science and statistical engineering would help considerably in this effort.

    1 Star2 Stars3 Stars4 Stars5 Stars (1 votes, average: 4.00 out of 5)


    • Noela M Prasad said:

      Thank you for publishing this thought-provoking and inspiring article that validates the immense contribution of people who work off-stage and outside the spotlight.
      Especially commendable is voicing the call to statisticians to take up complex challenges that exist in the real world; I hope this call is heeded.

    • Victor Morin said:

      Roger, you are to be commended for examining the difficult problems facing our world and instead of bemoaning their complexity and the difficulties they raise, you address possible solutions. It is a positive message of hope that we sorely need. Too often we do not take the time to step-back and think more globally and it is talks like yours that make us do so. Thank you.

    • Karen said:

      Bjørn Lomborg is a Danish economist, not Swedish.

      I really like the concept of statistical engineering and enjoyed the discussion of it in this article.

    • Roger Hoerl said:

      My sincere apologies; Bjorn Lomborg is Danish, not Swedish as I mistakenly stated.

    • Zubair Taiwo said:

      Greatest Statisticians! Fact from reliable figure. This article has been a comprehensive one that really emphasize on important of Statistics to global development. Yes we must answer the clarion call and provide the best possible result and decision making to this challenges through our data.*** I need sponsors.