ADVICE FROM THE LADY IN BLACK
Although male theoreticians dominated the development of statistical methods in the early years of the twentieth century, by the time I joined the profession in the 1960s, there were many women in prominent places. This was particularly true in industry and government. Judith Goldberg at American Cyanamid and Paula Norwood at Johnson Pharmaceuticals have headed up statistics departments at pharmaceutical companies. Mavis Carroll was in charge of the mathematical and statistical services division of General Foods. In Washington, D.C., women have been in charge of the Census Bureau, the Bureau of Labor Statistics, and the National Center for Health Statistics, among others. This has also been true in the United Kingdom and continental Europe. In chapter 19, we looked at the roles some ot these women have played in the development of statistical methodology.
There is nothing typical about the experiences of women who have made a name for themselves in statistics. All of them are remarkable individuals whose individual developments and accomplishments are unique. One cannot pick a representative woman in statistics, any more than one can find a representative man in statistics. It would be interesting, however, to examine the career of one woman, who rose to prominence in both industry and government. Stella Cunliffe of Great Britain was the first woman to be named president of the Royal Statistical Society. Much of this chapter is taken from the annual address of the president, which she presented before the society on November 12, 1975.
Those who have known or worked with Stella Cunliffe attest to her broad good humor, her keen common sense, and her ability to reduce the most complicated mathematical models to understandable terms for the scientists with whom she collaborated. Much of this comes out in her address. The address is a plea to the members of the Royal Statistical Society to spend less time developing abstract theory and more time collaborating with scientists from other fields. For example, she wrote: “It is no use, as statisticians, our being sniffy about the slapdash methods of many sociologists unless we are prepared to try to guide them into more scientifically acceptable thought. To do this there must be interaction between us.” She makes frequent use of examples where the unexpected has occurred in the process of running an experiment. “Barley trials, even on a well-organized research station, could be knocked for six by some fool of a tractor driver hurrying home to his tea via a short cut across the plot.”
Stella Cunliffe studied statistics at the London School of Economics in the late 1930s. It was an exciting time to be there. Many of the students and some of the faculty had volunteered to serve in the Spanish Civil War against the fascists. Prominent economists, mathematicians, and other scientists who had escaped Nazi Germany were given temporary positions at the school. When she emerged from school with her degree, the entire world was still suffering from the Great Depression. The only job she could find was with the Danish Bacon Company, “where the use of mathematical statistics was minimal and I, as a statistician, in particular a female
statistician, was looked upon as something very odd.” With the coming of World War II, Cunliffe became involved in food allocation problems, where her mathematical skills proved very useful.
For two years after the war, she volunteered to help the relief work in devastated Europe. She was one of the first to arrive in Rotterdam, in the Netherlands, while the German army was still surrendering and where the civilian population was starving. She moved on to help with the concentration camp victims at Bergen-Belsen soon after liberation. She finished her efforts working in the displaced-persons camps in the British zone of occupation. Cunliffe returned from her voluntary work penniless and was offered two jobs. One of them was at the Ministry of Food, where she would work in the “oils and fats” department. The other job was at the Guinness Brewing Company, which she accepted. Recall that William Sealy Gosset, who published under the pseudonym of “Student,” had founded the statistical department at Guinness. Stella Cunliffe arrived there about ten years after Gosset’s death, but his influence was still very strong at Guinness, where his reputation was revered and the experimental discipline he had created dominated its scientific work.
The Guinness workers believed in their product and the experimentation that was used constantly to improve it. They
never stopped experimenting to try to produce the product as a constant one, from varying raw materials, varying because of weather, soil, varieties of hops and barley, as economically as possible. They were arrogant about their product and, as may or may not be known, there was no advertising until 1929 because of the attitude—still endemic when I left—that Guinness is the best beer available, it does not need advertising as its quality will sell it, and those who do not drink it are to be sympathized with rather than advertised to!
Cunliffe described her first days at Guinness:
On arrival at the Dublin Brewery for “training,” having led, as I had in Germany, a free and in many ways exciting life, I appeared one morning before the supervisor of the “Ladies Staff” at the Dublin Brewery. She was a forbidding-looking lady, all in black, with small pieces of lace at her throat, held up by whalebones … . She impressed on me what a privilege it was to have been chosen to work for Guinness, and reminded me that I was expected to wear stockings and a hat and, if I was lucky enough to meet one of the chosen race known as “brewers” in the corridor, on no account was I to recognize him, but should lower my eyes until he had passed.
Thus the position of women in the hierarchal world of the Guinness Brewing Company in 1946.
Cunliffe soon proved her worth at Guinness and became deeply involved in the agricultural experiments in Ireland. She was not content to stay at her desk and analyze the data sent to her by the field scientists. She went out to the field, to see for herself what was happening. (Any new statistician would do well to follow her example. It is amazing how often the description of an experiment as relayed by somebody several layers above the laboratory workers does not agree with what has actually happened.)
Many is the damp and chill morning that found me at 7 A.M., shivering and hungry in a hop garden, actually taking part in some vital experiment. I use the word “vital” deliberately because unless the experiment is accepted as vital by the statistician, so that the enthusiasm of the experimenter is shared by the statistician, I submit that his contribution to the work
is less than optimum. One of the main problems as statisticians is that we have to be flexible: We have to be prepared to switch from helping a microbiologist in the production of a new strain of yeast; to helping an agriculturalist to assess the dung-producing qualities resulting from the intake of particular cattle feeds; to discussing with a virologist the production of antibodies to Newcastle disease; to helping a medical officer assess the effects on health of dust in malt stores; to advising an engineer about his experiments involving a mass-produced article moving along a conveyor belt; to applying queuing theory to the canteen; or to helping a sociologist test his theories about group behavior.
This list of types of collaboration is typical of the work of a statistician in industry. In my own experience, I have had interactions with chemists, pharmacologists, toxicologists, economists, clinicians, and management (for whom we developed operations research models for decision making). This is one of the things that make the work day of a statistician fascinating. The methods of mathematical statistics are ubiquitous, and the statistician, as the expert in mathematical modeling, is able to collaborate in almost every area of activity.
In her address, Stella Cunliffe muses on that greatest source of variability—Homo sapiens:
It was my delight to be responsible for much of the tasting and drinking experiments that are an obvious part of the development of that delicious liquid—Guinness. It was in connection with this that I began to realize how impossible it is to find human beings
without biases, without prejudices, and without the delightful idiosyncrasies which make them so fascinating … . We all have prejudices about certain numbers, letters, or colours, and all of us are very superstitious. We all behave irrationally. I well remember an expensive experiment set up to discover the temperature at which beer was preferred. This involved subjects sitting in rooms at various temperatures, drinking beers at various other temperatures. Little men in white coats ran up and downstairs with beer in buckets of water at varying temperatures, thermometers abounded and an air of bustle prevailed. The beers were identified by coloured crown seals, and the only clear-cut result of this experiment … was that our drinking panel showed that the only thing that mattered to them was the colour of the crown seals and that they did not like beer with yellow crown seals.
She describes an analysis of capacity of small beer casks. The casks were handmade and their capacity was measured to determine if they were of appropriate size. The woman who measured them had to weigh the empty cask, fill it with water, and weigh the full cask. If the cask differed from its proper size by being more than three pints below or more than seven pints above, it was returned for modification. As part of the ongoing process of quality control, the statisticians kept track of the filling size of the casks and which ones were discarded. On examining the graph of filling sizes, Cunliffe realized that there was an unusually high number of casks that were just barely making it, and an unusually low number that were just outside the limits. They examined the working conditions of the woman who weighed the casks. She was required to throw a discarded cask onto a high pile and place an accepted cask onto a conveyor belt. At Cunliffe’s suggestion, her weighing
position was put on top of the bin for the discarded casks. Then all she had to do was kick the rejected cask down into the bin. The excess of casks just barely making it disappeared.
Stella Cunliffe rose to head the statistical department at Guinness. In 1970, she was hired away by the Research Unit of the British Home Office, which supervises police forces, criminal courts, and prisons.
This unit, when I joined it, was concerned mainly with criminological problems and I plunged … straight from the rather precise, carefully designed, thoroughly analyzable work that I had been doing at Guinness into what I can only describe as the airy-fairy world of the sociologist and, if I dare say it, sometimes of the psychologist … . I am in no way decrying the ability of the researchers in the Home Office Research Unit … . However, it came as a shock to me that those principles of setting up a null hypothesis, of careful experimental design, of adequate sampling, of meticulous statistical analysis, and of detailed assessment of results, with which I had worked for so long, appeared to be much less rigorously applied or accepted in sociological fields.
A great deal of the “research” in criminology was run by accumulating data over time and examining it for the possible effects of public policy. One of these analyses had compared the length of sentence given to adult male prisoners versus the percentage of those men who were reconvicted within two years of discharge. The results clearly showed that the prisoners with short sentences had a much higher rate of recidivism. This was taken as proof that long sentences took “habitual” criminals off the streets.
Cunliffe was not satisfied with a simple table of rates of recidivism versus sentence length. She wanted to look carefully at the raw data behind that table. The strong apparent relationship was
due, in large part, to the high recidivism rate among prisoners given sentences of three months or less. But, upon careful examination, almost all these prisoners were “the many old, pathetic, sad and mad people [who] end up in our prisons because the mental hospitals will not take them. They represent a brigade that goes round and round.” In fact, because of their frequent incarceration, the same people kept showing up again and again but were being counted as different prisoners when the table was constructed. The rest of the apparent effect of long sentences on recidivism occurred at the other end of the table, where prisoners with sentences of ten years or more had less than a 15 percent rate. “There is a big age factor in this too,” she wrote, “a big environmental factor and a big offense one. Large frauds and forgeries tend to attract long sentences—but somebody who has committed a major fraud seldom commits another.” Thus, upon her adjusting the table for the two extreme anomalies, the apparent relationship between sentence length and recidivism disappeared.
As she wrote:
I opine that even the so-called “dull old Home-Office statistics” are fascinating … . It seems to me that one of the statistician’s jobs is to look at figures, to query why they look like they do … . I am being very simpleminded tonight, but I think it is our job to suggest that figures are interesting—and, if the person to whom we say this looks bored, then we have either put it across badly or the figures are not interesting. I suggest that my statistics in the Home Office are not boring.
She decried the tendency for government officials to make decisions without careful examination of the data available:
I do not think this is the fault of the sociologist, the social engineer, the planner … but it must often be laid very
firmly to the door of the statistician. We have not learned to serve those disciplines that are less scientific than we might like, and we have not therefore been accepted as people who can help them to further knowledge … . The strength of the statistician in applied fields, as experienced by me … lay in his or her ability to persuade other people to formulate questions which needed answering; to consider whether these questions could be answered with the tools available to the experimenter; to help him set up suitable null hypotheses; to apply rigid disciplines of design to the experiments.
In my own experience, the attempt to formulate a problem in terms of a mathematical model forces the scientist into understanding what question is really being posed. A careful examination of resources available often produces the conclusion that it is not possible to answer that question with those resources. I think that some of my major contributions as a statistician were when I discouraged others from attempting an experiment that was doomed to failure for lack of adequate resources. For instance, in clinical research, when the medical question posed will require a study involving hundreds of thousands of patients, it is time to reconsider whether that question is worth answering.
Stella Cunliffe emphasized the hard work of making statistical analyses useful. She was disdainful of elaborate mathematics for the sake of mathematics and decried mathematical models that are
all the imagination and lack of reality … lots of string, interesting side-pieces, plenty of amusement, brilliant of
concept, but also the same lack of robustness and of reality. The delight in elegance, often at the expense of practicality, appears to me, if I dare say so, to be rather a male attribute … . We statisticians are educated to calculate … with mathematical precision … [but] we are not good at … persuading the uninitiated that our findings are worth heeding. We shall not succeed in so doing if we solemnly quote “p less than 0.001” to an incomprehending man or woman; we must explain our findings in their language and develop the powers of persuasion.
Without a hat, and refusing to make herself meekly subordinate to the master brewers, Stella Cunliffe flew into the world of statistics, jauntily indulging her lively curiosity and criticizing the professors of mathematical statistics who came to hear her speech. As of this writing, she could still be found attending meetings of the Royal Statistical Society and skewering mathematical pretensions with her tart wit.