Rating: 
Summary: Best book for the general reader on statistics
Review: Most people don't realize that the very notion of proof, at least in the field of medicine, did not exist until 1934, when the founder of modern statistics, R.A. Fisher, invented it. He would undoubtedly have some scathing remarks on what currently passes for proof for new medical treatments. You'll read about all the great statisticians of the 20th century, many of whom fled the Nazi's, or the Russians, and wound up in the United States. One accomplished American statistician was laid off by Department of Agriculture bureaucrats in the great Depression and could only find a job in the Soviet Union under Stalin. What a great story! I met some of these guys at Stanford when I was getting my Masters Degree in Statistics in the 1970s. While they sometimes can be boring on the surface, underneath lurks a passion for reality rarely found in more superficially interesting folk. I used the text of Gumbel on how to compute the probability of a 100 year flood as the basis of my Ph.D. thesis on carcinogenesis at the University of Colorado School of Medicine Department of Biometrics in the 1980s. As a well rounded technologist, Gumbel also published a book on Four Years of Political Murder in 1922, followed by Causes of Political Murder in 1928, as a critique of the Nazis. When the Nazis came to power in 1933, he barely escaped Germany and had to hide out in Southern France. This is the best book of this type that I've read since Fermat's Enigma and it is best savored chapter by chapter over a cup of cappuccino in a Peets or Starbucks. A book for the general reader that every statistician should read!
Rating: 
Summary: Nice shot at a difficult topic
Review: Salsburg sets himself the rather difficult task of explaining how the ascension of statistics as a "valid" form of reasoning in the early 20th century has altered the course of science. And while "The Lady Tasting Tea" is enjoyable, it doesn't make for compelling page-turning either as philosophy of science or as biography. I can excuse Salsburg for not going into more depth mathematically, but I still felt like many of the concepts were dangled just out of reach. Although the role of statistics in science is the focus of the early part of the book, this historically fascinating and still relevant topic is never fully developed. Salsburg never really gets down to the subtle job of teasing apart determinism and non-determinism; deterministic theories and statistical ones; absolute truth and pragmatism; and cognitive agents, theories and reality. A deeper inquiry would've inevitably led to the philosophical context of logical positivism and empiricsm, with philosophers such as Popper, Russell, and later Quine struggling with statistics and its relation to induction as the underpinning of scientific theory.
The biographical sketches, which form the bulk of the book, provide a straight-line historical organization to the subject matter. Although the section on Pearson gets off to a good start, the later sketches are much weaker and read like encyclopedia briefs. It almost feels as if the author gave up halfway through the book when he saw the enormity of the task he had set for himself. By way of contrast, James Gleick managed to convey a real sense of the the interplay of theory, personality, publication and conferences in the much larger book "Genius", though having Richard Feynman as a subject made Gleick's job substantially easier. Paul Hoffman also painted a moving picture of Paul Erdos, the peripatetic mathematician in the shorter "The Man Who Loved Only Numbers". Thus writing a compelling book about science and scientists clearly isn't impossible. In Salsburg's defense, perhaps there just weren't as many colorful personalities in statistics as in physics or mathematics; one gets the impression of a rather dour bunch on the whole.
Rating: 
Summary: What happened to Frank P. Ramsey?J M Keynes?
Review: Salsburg(S) does an excellent job discussing the historical development of the field of statistics in the 20th century.He has a way of writing that blends current statistical theory with the development of statistics over time from a historical perspective with the individuals who made it all happen,such as Neyman-Pearson and Sir Ronald Fisher.In this book he is close to Ian Hacking in the manner in which he weaves his story.This reviewer has a few quibbles.First,in S's discussion of the personalist(subjectivist)theory of probability,only de Finetti and Savage are covered.Since Frank Ramsey's 1922 and 1926 contributions to the subjective theory of probability,unfortunately combined with error filled critiques of John Maynard Keynes's logical theory of probability,were published BEFORE the work of de Finetti and Savage,he definitely deserved to have a prominent place in any book dealing with the history of probability and statistics.Second,there are a number of errors made in the all to brief discussion of Keynes and his logical theory of probability in his 1921 book,A Treatise on Probability(TP).Contrary to S(p.112,p.305),Keynes never received a doctorate in philosophy for writing the TP because the TP is not a doctoral dissertation.The TP was a thesis submitted for a fellowship, successfully, in 1909 at Cambridge.Keynes added a Part V to his thesis in the period from 1910-1914 to complete his TP.S commits another error when he chacterizes Keynesian economic policy as the manipulation of monetary policy.It is the manipulation of both fiscal and monetary policy.Finally,Keynes's probabilities are primarily intervals with a lower and an upper bound,not ordinal rankings as suggested by S.S's flawed appraisel involves a failure to translate Keynes's definition of the term "nonnumerical",which means"not by a single numeral but by two numerals".Finally,S is in too much of a hurry to take the side of Neyman,a deductivist, in his debates with Fisher,an inductivist,about significance levels(p-values) and confidence intervals.Neyman's justification for confidence intervals is badly flawed.It essentially boils down to an arbitrary "act of will" on the part of the researcher.Fisher,who was well acquanted with Keynes's logical theory of probability,realized that Neyman's "reasoning" was actually an evasion.Unfortunately,Fisher never was clear about his reservations .Fisher simply needed to come right out and say that a 95% confidence interval means that the researcher is 95% confident that the particular parameter,say the mean,lies in that interval.Of course,this conclusion follows from the proportional syllogism,which is part of the logical theory of probability.Neyman,who was a frequentist,ends up in a quagmire of his own creation because he did not want to allow any "inductive" concepts into his theory.
Rating:
Summary: Wonderfully well written, entertaining, and informative
Review: The intense media attention given to the proof of Fermat's Last Theorem a few years ago was followed by the publication of many books on mathematics for non-mathematicians. Dr. Salsburg's book is arguably among the best of them. It has many interesting and illuminating anecdotes about the most influential statisticians in the early 20th century, which is when the Statistical Revolution (as aptly called by the author) took place. Important developments are clearly explained in their historical context, and their implications for current (i.e., 21st century) scientific research are given. The student of Statistics will get to know the people behind the names mentioned in the textbooks. The book is non-technical and written for the general public, but as a statistician myself I can say that I was no less than delighted reading it. In fact, two chapters (on probit and sample selection) deal with concepts I'm using in an epidemiological manuscript!
Rating:
Summary: great look at statistics in the 20th Century
Review: The Lady Tasting Tea is a new book by David Salsburg (a Ph.D. mathematical statistician, who recently retired from Pfizer Pharmaceuticals in Connecticut). The title of the book is taken from the famous example that R. A. Fisher used in his book "The Design of Experiments" to express the ideas and principles of statistical design to answer research questions. The subtitle "How Statistics Revolutionized Science in the Twentieth Century" really tells what the book is about. The author relates the statistical developments of the 20th Century through descriptions of the famous statisticians and the problems they studied. The author conveys this from the perspective of a statistician with good theoretical training and much experience in academia and industry. He is a fellow of the American Statistical Association and a retired Senior Research Fellow from Pfizer has published three technical books and over 50 journal articles and has taught statistics at various universities including the Harvard School of Public Health, the University of Connecticut and the University of Pennsylvania. This book is written in layman's terms and is intended for scientists and medical researchers as well as for statistician who are interested in the history of statistics. It just was published in early 2001. On the back-cover there are glowing words of praise from the epidemiologist Alvan Feinstein and from statisticians Barbara Bailar and Brad Efron. After reading their comments I decided to buy it and I found it difficult to put down. Salsburg has met and interacted with many of the statisticians in the book and provides an interesting perspective and discussion of most of the important topics including those that head the agenda of the computer age and the 21st century. He discusses the life and work of many famous statisticians including Francis Galton, Karl Pearson, Egon Pearson, Jerzy Neyman, Abraham Wald, John Tukey, E. J. G. Pitman, Ed Deming, R. A. Fisher, George Box, David Cox, Gertrude Cox, Emil Gumbel, L. H. C. Tippett, Stella Cunliffe, Florence Nightingale David, William Sealy Gosset, Frank Wilcoxon, I. J. Good, Harold Hotelling, Morris Hansen, William Cochran, Persi Diaconis, Brad Efron, Paul Levy, Jerry Cornfield, Samuel Wilks, Andrei Kolmogorov, Guido Castelnuovo, Francesco Cantelli and Chester Bliss. Many other probabilists and statisticians are also mentioned including David Blackwell, Joseph Berkson, Herman Chernoff, Stephen Fienberg, William Madow, Nathan Mantel, Odd Aalen, Fred Mosteller, Jimmie Savage, Evelyn Fix, William Feller, Bruno deFinetti, Richard Savage, Erich Lehmann (first name mispelled), Corrado Gini, G. U. Yule, Manny Parzen, Walter Shewhart, Stephen Stigler, Nancy Mann, S. N. Roy, C. R. Rao, P. C. Mahalanobis, N. V. Smirnov, Jaroslav Hajek and Don Rubin among others. The final chapter "The Idol with Feet of Clay" is philosophical in nature but deals with the important fact that in spite of the widespread and valuable use of the statistical methodology that was primarily created in the past century, the foundations of statistical inference and probability are still on shaky ground. I think there is a lot of important information in this book that relates to pharmaceutical trials, including the important discussion of intention to treat, the role of epidemiology (especially retrospective case-control studies and observational studies), use of martingale methods in survival analysis, exploratory data analysis, p-values, Bayesian models, non-parametric methods, bootstrap, hypothesis tests and confidence intervals. This relates very much to my current work but the topics discussed touch all areas of science including, engineering in aerospace and manufacturing, agricultural studies, general medical research, astronomy, physics, chemistry, government (Department of Labor, Department of Commerce, Department of Energy etc.), educational testing, marketing and economics. I think this is a great book for MDs, medical researchers and clinicians too! It will be a good book to read for anyone involved in scientific endeavors. As a statistician I find a great deal of value in reviewing the key ideas and philosophy of the great statisticians of the 20th Century. I also have gained new insight from Salsburg. He has given these topics a great deal of thought and has written eloquently about them. I have learned about some people that I knew nothing about like Stella Cunliffe and Guido Castelnuovo. It is also touching for me to hear about the work of my Stanford teachers, Persi Diaconis and Brad Efron and other statisticians that I have met or found influential. These personalities and many other lesser-known statisticians have influenced the field of statistics. The book includes a timeline that provides a list in chronological order of important events and the associated personalities in the history of statistics. It starts with the birth of Karl Pearson in 1857 and ends with the death of John Tukey in 2000. Salsburg also provides a nice bibliography that starts with an annotated section on books and papers accessible to readers who may not have strong mathematical training. The rest of the bibliography is subdivided as follows: (1) Collected works of prominent statisticians, (2)obituaries, reminiscences, and published conversations and (3) other books and article that were mentioned in this book. The book provides interesting reading for both statisticians and non-statisticians.
Rating:
Summary: Statistics humanized and triumphant
Review: The title refers to the story about the English lady who believed she could tell by tasting whether the milk had been added to the tea or the tea added to the milk. We find out here that apparently she could. At least in the small sample of cases recorded, she "identified every single one of the cups correctly." (p. 8) The question--and this is the question that statisticians are forever trying to answer--is, was the result significant? Or how much faith should we put in such a result? What is the probability that such a result comes to us by chance rather than by causation? Did she simply guess right ten times in a row? Or, more saliently, how many times would she have to guess right before you'd be a believer? Or, more rigorously, how many times out of how many trials would she have to guess right before we can be confident that she isn't just guessing? Statistics then is a way of understanding and appreciating events without reference to causation. How cigarette smoking causes lung cancer is not exactly known. The fact that cigarette smoking does indeed cause lung cancer is demonstrated by a clear statistical correlation between smoking and the instance of lung cancer. But is a statistical correlation proof? Salsburg's very readable book is a narrative about the mathematicians who have tried to answer this and other statistical questions. The emphasis is on the mathematicians themselves, not on their mathematics. Indeed, following a time-honored "rule" in the book publishing business, a rule that insists that you lose "x" number of readers for every mathematical formula that appears on your pages, Salsburg has elected to use a grand total of zero. I was a little disconcerted about this. To encounter Bayes's theorem or any number of other statistical ideas and see not a single formula or mathematical expression was to me like reading a joke book without any jokes in it. But for those who have heard the jokes and are only interested in the joke tellers and their problems, this is indeed a fascinating book. It is ironic that this "non-mathematical" book is probably best appreciated by those with some experience with statistics. Such readers I suspect will be quite pleased to read about the lives of such greats in statistical theory and methods as Karl Pearson, R. A. Fisher, William Sealy "Student" Gosset, John Tukey, etc. Salsburg focuses on the problems that the individual mathematicians encountered and the solutions they came up with. Here's an example of how Salsburg does this neat trick of talking about mathematics without using any mathematics. He asks, "What is the central limit theorem?" (p. 84) and answers thusly: "The averages of large collections of numbers have a statistical distribution. The central limit theorem states that this distribution can be approximated by the normal probability distribution regardless of where the initial data came from. The normal probability distribution is the same as Laplace's error function. It is sometimes called the Gaussian distribution. It has been described loosely in popular works as the bell-shaped curve." Perhaps this does work for a lot of people, but I think this book would be improved if there were an appendix with a list of ideas, presented in mathematical form. For a new edition, Salsburg might want to do something like that. Then this interesting book would also be a work of reference. My favorite method learned here is on page 236. Salsburg describes how John Tukey believes one should tally. Instead of making vertical lines and crossing every fifth one (which is what I have done for decades) Tukey recommends "a ten-mark tally. You first mark four dots to make the corners of a box. You then connect the dots with four lines, completing the box. Finally, you make two diagonal marks forming a cross within the box." That statistical ideas are inexorably tied up with the ideas of probability is explored in the final chapter of the book, "The Idol with Feet of Clay." Salsburg observes, along with Thomas Kuhn, that we are forever describing reality with "a model...that appears to fit the data," but as the data accumulates our model "begins to require modifications." (p. 293) Reality in this sense is the postulated "universe" of the statistician, and our experiences and "laws" the result of "samplings" of that universe. Salsburg, citing L. Jonathan Cohen, goes on to recall Seymour Kyberg's "lottery paradox" which makes it clear that statistical/probabilistic "proofs" run into logical problems. He then asks if we really understand probability. He recalls the notion of "personal probability" (something I used to call "psychological probability") in which we appreciate the probability of something happening in terms of what effect it might have on us personally. Thus a small chance of getting something exceeding important to us (such as winning the lottery) might be worth paying more for the ticket than it is objectively worth. Salsburg concludes that we really do not understand probability except in the grossest sense (e.g., "50/50" or "almost certain"). Then he asks, does it matter? His answer suggests quantum mechanics in which we work with probabilities without any pretense of grasping underlying "laws." Salsburg ends the book with a yearning for a new paradigm without feet of clay. I suspect he has in mind the undeniable and always troubling fact that the best that can ever be said about a sampling is that it has a certain probability of being an accurate reflection of the entire universe. However, my guess is that we will continue to have to be satisfied with "only" probabilistic knowledge; indeed that knowledge itself will always be subject to some degree of doubt. I might even conjecture that all real world knowledge, yearn as we might for certainty, is probabilistic.
Rating:
Summary: strictly for non-Bayesians
Review: This book has many good qualities. It is easy to read, and I enjoyed reading it. It is also cheap and light in weight, with short chapters, so I read most of it traveling on the subway. The historical anecdotes about famous statisticians are interesting and enliven the book. But it has two drawbacks.
First, its references are not really up to modern standards. If something catches your eye and you want to follow it up, the book does not make it easy for you. There are several pages of references, but they are not linked to the text and they are not arranged by topic.
The second drawback will probably pass most readers by, but is more serious. David Salsburg appears to be a resolute non-Bayesian. He mentions some Bayesian ideas (one chapter out of 29 is the "Bayesian Heresy"), but he is clearly unsympathetic. The problem about this is that he manages to miss entirely the fascinating story of how some demonstrably wrong ideas ("classical statistics") took over from Bayesian statistics in the early twentieth century and have held sway ever since. In many ways it is classic Kuhn - we are waiting for the "classical statistics" guys to die off. Like all stories about science there are many fascinating subplots, but Salsburg manages to miss it all.
He also, of course, helps to educate the lay reader (at whom the book is aimed) in some seriously wrong ideas.
Rating:
Summary: Anecdotal history of Twentieth Century statistics
Review: This book introduces the personalities behind the names found on the famous statistical works of the Twentieth Century. Beginning with Francis Galton, the founder of the journal Biometrika and the discoverer of the uniqueness of fingerprints, Salsburg outlines the major developments and developers of modern statistics. In order to make the book accessible to general readers, he strenuously avoids mathematical formulas or charts, keeping his discussion focused on the people behind the math. He relates such tales as the origin of "Student t-tests", which go back to William Sealy Gossett, a statistician employed by the Guiness Brewing Company, who was forbidden by his company to publish his work, hence his use of the pseudonym "Student". The text is organized into many short chapters, each only a few pages long. At the end of the book is a timeline, covering the publication dates of key papers in statistics and their authors, followed by an annotated bibliography of suggested works for further study and a list of materials used in the book. There is also an index that includes names of people and institutions as well as general statistical topics.
I picked up the book because I was intrigued by the sub-title: "How Statistics Revolutionized Science in the Twentieth Century." Unfortunately, the book has very little about this topic, and probably much more could be said- -it certainly would make for an interesting volume. The numerous stories about the people behind the developments in statistics are quite interesting, nevertheless. Unfortunately, Salsburg goes a bit too far in avoidance of math. He describes statistical topics in a very general fashion, so general in fact, that readers who don't know statistics are left completely in the dark. If he had only added a graph here and there to demonstrate the topics visually, interested general readers might gain a better sense of what each statistical personality accomplished. He also has a habit of laying out the details of interesting experiments in such fields as medicine or agronomy which led to the development of new statistical approaches. But then he leaves us hanging, not following up with the results of the experiments and the scientific facts that were learned through using the statistics.
Nevertheless, the book is quite engaging, and I've found it has at least sensitized me to importance of statistics (without actually teaching me how to do any statistics). It would be very valuable reading for statistics students, enabling them to get to know the people who wrote the famous papers in their field and learn about the circumstances that led to their discoveries.
Rating:
Summary: Wonderful book for both mathematical and non-mathematical
Review: This book is a wonderful depiction of the history of Statistics and its great contributors. Dr. Salsburg conveys the stories of the great minds of the statistical world in an insightful and interesting way.
Rating:
Summary: Engaging history of statistics, but illiterate on science
Review: This book is almost wonderful. It presents an account of the history of statistics that I almost couldn't put down. While other reviewers have complained about the lack of detail in the descriptions of statistical principles and techniques, what this book does have, which I haven't seen anywhere else is an informal account that guides the reader to the original literature. Salsburg is a wonderful writer. The book focuses on biographical anecdotes, some of them from personal interactions Salsburg had with his subjects, which follow the development of 20th century statistics. Salsburg's love of his subject is contageous and I repeatedly found myself staying up later than was good for me to read another chapter. I would guess that much of the book might be hard to follow for someone not familiar with statistics, but if you know what an F test or a t test is, this will give you a new appreciation of how they were invented and may well inspire you to drop by the library and read the original papers. The weakness of this book is Salsburg's tendency to pontificate about areas of which he clearly knows next to nothing. For some bizarre reason, he dismisses chaos theory (something that certainly has been oversold to the public) saying that this discipline has no more rigorous methods than asserting the subjective similarity of Poincare maps and that there is no experimental evidence that Nature is chaotic. Apparently, Salsburg could not be bothered to glance through even the most basic literature of the field, such as the experimental confirmation of universality in period doubling, the rigorous quantitative application of the Lorenz equations (which he dismisses) to experimental data from bistable lasers, or techniques such as Grassberger and Procaccia's of rigorously determining the properties of Poincare maps. This stuff has been in the mainstream scientific literature for over 20 years, so Salsburg has no excuse to make pronouncements without learning anything about it. In fact, Salsburg seems to use chaos as a stalking horse for an attack on determinism in physical science, but he misses the notion that Lyopunov expansion and much of KAM theory can be expressed just fine in the statistical notation that Salsburg is fond of. The burgeoning field of quantum chaology has matured enough in the for the last 10 years or so it has been possible to investigate the chaotic behavior of statistical systems rigorously and meaningfully, but Salsburg only advances his nondeterministic view of the physical world in back-handed asides, and never presents it to us straightforwardly for inspection. Still, if you forgive Salsburg his ignorance of science, he clearly is a master of the mathematical statistics he presents and this is an engaging book that I am very glad to have in my library and which I am heartily recommending to my friends.
|
