Sharon Bertsch McGrayneReseñas

Unfortunately, very disappointing. I didn't really need to hear 10 different stories that said Bayes could solve x, without any of them really explaining how.

Bayes theory is cute. Pop nonfiction math books seem incapable of being patronizing on one extreme or invoking their math theorem as an abstract magical spell on the other. I prefer the later, which is what this is. How did we find Russian submarines? We cast Bayes at them. Sometimes, even as someone very familiar with Bayes theorem I found these invocations impossible to understand what was literally happening, but overall, this is an easy and mathy read. 3.5 stars.

Bayes is a statistical technique for estimating probability that starts off with a guess as an initial condition. This guess has brought it a lot of flack since it was invented in about 1760 from scientists and mathematicians who find the guess unscientific. For most of the 250 years since then it has been niche technique, not quite acceptable in polite mathematical circles, if not provoking outright hostility. However, its influence has grown hugely since the advent of computers which make the enormous calculations it requires practical. Surprisingly gripping yarn and very approachable. Recommended.

I came to this book hoping to understand what the heck scientists mean when they say they use a Bayesian approach or Bayesian statistical analysis, but without having to decipher too many formulas or greek letters. However, the book may have erred too much on the side of popular nonfiction; I'm surprised that after reading, I only have a slightly better understanding of Bayes than before.

But I can't exactly fault the author. I doubt there is much market for a popular explanation of Bayesian statistics, and it is a more intriguing and sellable book to chart the origins and many different applications of Bayes. There is this nagging feeling on my part that the book lacks grounding on some level...so many things are described as Bayesian, but the nitty gritty details of each problem that would help us really see them as such are missing.

The book also seems to end rather abruptly and without conclusion. We finally come to the flourishing of Bayesianism in the latter portion of the 20th century, and the advent of more powerful computers that help crunch the numbers, but the author treats it all rather cursorily, passing quickly from example to example.

Still, I found the book fairly interesting. I would also recommend David Salsburg's The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century for those looking for another well-written popular take on statistics.

I loved this book. I didn't love many of the experiences she wrote about, but I did love the way she wrote about them, and what she did about them.½

This is a very important and enraging book. The author has some fascinating stories to tell about her life in science; she’s done a lot of good for the world and has worked to bring more recognition to women as well. The enraging part as always is the men and what they’ve done to actively reduce the contributions of women in science. The author has some beneficial yet depressing advice for women in the science field at the end of the book. God I’m grumpier than usual now...½

With The Theory That Would Not Die Sharon Bertsch Mcgrayne takes the reader on an adventurous romp through the history of Bayesian Analysis. From the initial founding by Thomas Bayes to its refinement by Robert Price and perfection by that Master of Mathematics Pierre-Simon Laplace, we explore its uses and abuses by many different people.

Having been shelved several times, it is surprising that one can even trace Bayes Rule to a single person, but it is the case here. Although the Frequentists of the past would have several complaints about its use in many different situations, there is no denying the utility of Bayes’ Rule. The primary reason against its use is the fact that initially, you have to make a guess on the probability of something. However, the power of Bayes is the fact that you are able to refine your guess as new information comes in.

As I mentioned, Bayes Rule was invented or first used by a man named Thomas Bayes. He invented it to find the probability of something that was unknown with limited information. Since he was a devout man and a theologian, that practical use turned out to be some sort of proof of God. After Bayes’ death, Price worked to make it more rigorous and mathematical. After that, Laplace developed it independently since he was such a genius, but acknowledged Bayes once he found out about him. Once Laplace died, no one used Bayes for a while since Laplace made it somewhat confusing. It didn’t help that Laplace was such a genius that he could skip several important steps when doing the math, preferring instead to put things like “this much should be obvious.”

In the realm of Statistics, Bayes was maligned and in many ways vilified. People made it their lifelong quest to call Bayes Theorem ridiculous and silly. This was mainly because you had to make an initial guess and that was not considered scientific. Eventually, it was used for all sorts of things. The main thing it was used for was insurance. If you were an actuary in the 1930s you had to figure out how to apply a probability to something that had not happened yet. So they used Bayesian Analysis to figure out insurance tables. As I said, the beauty of Bayes is that it grows more and more accurate as data accumulates. With Birth and Death records, you could find the probability of a ton of things. World War II was another time for Bayes to shine, but it happened with utmost secrecy. This is because it was used to break the German Enigma codes and end the war sooner.

So in the more modern times following WWII, it has been used for many different things. For instance, back in the 1950s or so, it was discovered that incidents of Lung Cancer and Heart Disease were on the rise. Using Bayes Rule, Dr Jerome Cornfield connected this rise to smoking cigarettes. Armed with Bayes Rule, he made many other earthshaking discoveries. Dr Cornfield was the one who made the connection between birth defects and thalidomide, the anti-nausea drug. In the 1960s, more people were driving and more airplanes were flying. So someone asked an actuary “what are the chances of two planes colliding midair?” Again, with the traditional realm of Frequentist statistics that question wouldn’t have made any sense. You would have to have something that happened in order to count it and get statistics from that. However, no one wants to frequently crash a plane into another plane to get data, so another method was used.

In any case, this book was really interesting. I have the version that contains a new preface, an epilogue, and a series of case studies. For instance, if you are a woman and get a positive mammogram, what are the chances you have Breast Cancer given that you have no family history or any other negative signs? The chances are pretty low actually. As it turns out, since that particular cancer is rare, it is more probable that you received a false positive.

Bayes really deserves a better book than this one. I don't think it's particularly well-written, and it makes the subject matter appear drier than it really is. I think there's a good book in the subject matter, but this one isn't it.

Students struggling in an introductory chemistry course with the difficulties of the subject’s fundamentals could be forgiven for thinking no individuals that they’d care to know were involved. Chemistry can seem a sort of hidden subterranean conjuring governed by obscure wizards with bent bodies and crumpled, cranky souls whose products, as an 18th century visitor to Liverpool noted, were “pills, coal, glass, chemicals, cripples, millionaires, and paupers.” Is that fair?

Well, mostly no. In Sharon Bertsch McGrayne’s Prometheans in the Lab we meet some important actors in the story. They’re worth hearing about. Readers already averse to chemistry’s technicalities will find a few chemical structures and equations with which to grapple but these aren’t often an impediment.

The author has assembled a diverse group (if one is permitted to call an all-male group “diverse”) who reveal a spectrum of character to admire or decry. In the final chapter, we find a hero in Clair C. Patterson, whose work led famed novelist Saul Bellow to nominate him for the Nobel Prize. We are brought to appreciate the social contexts in which these men worked. The stories present the complexities of attempting that which benefits us at risk of damaging us too. The author doesn’t much pursue explicitly how to reconcile such diverging effects or how to value one act over another and perhaps it’d be a stronger book had she attended to it more. That could, however, detract from the narratives she chose to tell, narratives which interested and surprised me. These stories make Prometheans in the Lab a fine contribution among books discussing chemistry.

Bayes really deserves a better book than this one. I don't think it's particularly well-written, and it makes the subject matter appear drier than it really is. I think there's a good book in the subject matter, but this one isn't it.

Lacks sufficient mathematical presentation to make the history meaningful. The history is fun and interesting but frustrating, because the mathematics is always vague. There is a bibliography at the end of the book, but you would have to search it fairly carefully to find the stuff that would make the mathematics clearer.

Unfortunately a little light on the maths - Bayes famous equation barely makes the cut. However, it is a good and fascinating read, even if it's a little too detailed. This must be the authoritative history of the subject. And it's tweaked my interest enough to build some spreadsheet models and download some of the key papers it refers to. I found the descriptions of the optimal search strategies (e.g. for submarines, or lost nuclear bombs) totally fascinating. Thoroughly recomended.

Even if you failed math in college this book is good. Basically it's a history one statistical theorem. Put simply by updating intial your belief about something with objective new information , you get a new and improved belief. So simple but so controvrsial. The book is not easy to read in fact I was Googling some of the ideas, persons, places mentioned to understand what was going on but in the end it was worth all the time spent.

Sharon McGrayne is a very good and engaging writer. She has an interesting story to tell about the last 250 years of Bayesian thinking, how the theory has developed, and its many applications including how to price insurance, how to aim artillery, how to break the Enigma code, who wrote The Federalist Papers, how to find Russian nuclear subs, how to estimate the probability of a shuttle disaster, when to do various cancer screenings, whether cigarette smoking is harmful, etc. She also has a great set of characters, a parade of statisticians who are more colorful than I could have imagined, from the pioneers of Bayes, Price and Laplace to most recent statisticians like Cornfield, Tukey and Mosteller.

But, the book is deeply flawed and disappointing because it does so little to actually explain Bayes Theorem, how it was applied, how it led to different confusions than frequentism, and how the two have recently been theoretically synthesized. Most of this is not very complicated, one knows a decent amount already, but it would be more interesting to understand hot it was applied. Instead, the book concentrates much more on personality and the more surface descriptions rather than dwelling deeper and working out at least a few examples in more detail, both more of the theory from first principals but also better understanding what data and calculations various of her protagonists were using. Absent that, the book is often literally superficial.

Still, the book has a lot of upside -- but given that there is not exactly a huge selection of books covering this ground (unlike, say, quantum mechanics) to have this as nearly the sole choice is disappointing.

I was disappointed by this book. I had read a good review that played up this interesting view on the world of statistics and how the Bayes Rule had becoming a compelling tool for statistical analysis.

Perhaps it was my poor showing in statistics during my college days. Perhaps I wanted to make up for my mistake of signing up for a course like stats that had class on Friday afternoons. But I thought this book could be interesting, despite the somewhat dry core subject.

The book did little for me in helping to show how the rule worked in theory and in practice. It did a great job of showing the power of Bayes Rule as a problem solving tool. It may have lead to the victory in WWII for its use in cracking the German U-Boat enigma codes.

Maybe I had my expectations too high and maybe I was looking for more substance about the statistics tools themselves

Interesting subject, very poorly written½

A well written history of Bayes Theorem. I found the information fascinating and educational. This is one of the best books I have read in years. I strongly recommend this to anyone who has an interest in statistics, science, and history.

An enjoyable collection of biographies of Bayesians and their 'enemies'. Not much was made of the irony that neither Thomas Bayes nor Pierre Laplace would have been 'Bayesians'. The most 'practical' part of the book was the indirect pointer to the OpenBUGS software project, http://www.openbugs.info/w/. (I'd like to explore using it for making software project estimates ... even if it's already been done.) For me, one of the most interesting factoids was that Bayes came up with his approach as a response to David Hume's thoughts on the problem of induction; Bayes wanted to show that it was probable that God was the creator of the universe. Towards the end of the book, McGrayne mentions that some modern-day philosophers have carried on in that vein. All in all, it was good fun.

The book basically says that some people like using Bayesian statistics, and other people don't think that Bayesian statistics should be utilized. However, it did not discuss in any detail how to apply Bayesian statistics to any actual problems, nor does it give any numeric examples. Although there were many pages of references in the back, the book did not seem to make any points clearly.

I love Bayesian probability, and here's a book about it! Actually it's mostly about the people and the politics surrounding the Bayesians' fight against frequentists (who didn't believe in using the subjective probabilities that full-on Bayes/Laplace analysis requires). McGrayne tracks the various uses of Bayes to solve problems across multiple fields, from cryptography to finding lost submarines, but I really wished it had been mathier: I felt like a lot of times I was taking her word that Bayes made the problem at issue easier to solve than frequentism. Concededly, it can be super hard to explain this--I have had hour-long discussions with very smart people over the Monty Hall problem. But I wished she'd tried more; her explanation of Monte Carlo modeling was clear and easy to follow.½

In the realm of applications of mathematical probability and statistics, this is a good personality-oriented history of the Bayesian approach as philosophically (or ideologically or, as McGrayne unfortunately puts it, "religiously") opposed to the frequentist approach. The narration is, rightly enough, peppered with phrases such as "subjective probability", "equal priors", and "probability of causes", but their technical basis is not, in my opinion, adequately spelled out. There is nothing more detailed than a short appendix that uses the simple
P(A|B) = P(B|A)P(A) / P(B)
form of Bayes's Theorem (where A is a hypothesis, B is evidence, "P(...)" means "probability of", and "|" means "given") to solve a simple problem that does not involve any subjectivity. Readers who want to better understand how the controversial concepts in Bayesian practice relate to the underlying math will need to consult supplementary material. As a starting point, I'd very highly recommend the "excruciatingly gentle introduction" at http://yudkowsky.net/bayes/bayes.html, which uses the more elaborate
P(A|B) = P(B|A)P(A) / (P(B|A)P(A) + P(B|~A)P(~A))
form of the theorem. To thumbnail McGrayne's history itself: "Thomas Bayes had turned his back on his own creation; a quarter century later, Laplace glorified it. During the 1800s it was both employed and undermined. Derided during the early 1900s, it was used in desperate secrecy during the Second World War and afterward employed with both astonishing vigor and condescension." (p 176) "By 1978 the ... frequentists held an 'uneasy upper hand' over the Bayesians ..." (p 178) In the most recent decades, certain technical advances (such as Markov-chain Monte Carlo, Gibbs sampling, Kalman filters, and BUGS software) have promoted Bayesianism's acceptance and employment to the point where many people claim, "We're all Bayesians now." (p 235)

Sharon McGrayne is a very good and engaging writer. She has an interesting story to tell about the last 250 years of Bayesian thinking, how the theory has developed, and its many applications including how to price insurance, how to aim artillery, how to break the Enigma code, who wrote The Federalist Papers, how to find Russian nuclear subs, how to estimate the probability of a shuttle disaster, when to do various cancer screenings, whether cigarette smoking is harmful, etc. She also has a great set of characters, a parade of statisticians who are more colorful than I could have imagined, from the pioneers of Bayes, Price and Laplace to most recent statisticians like Cornfield, Tukey and Mosteller.

But, the book is deeply flawed and disappointing because it does so little to actually explain Bayes Theorem, how it was applied, how it led to different confusions than frequentism, and how the two have recently been theoretically synthesized. Most of this is not very complicated, one knows a decent amount already, but it would be more interesting to understand hot it was applied. Instead, the book concentrates much more on personality and the more surface descriptions rather than dwelling deeper and working out at least a few examples in more detail, both more of the theory from first principals but also better understanding what data and calculations various of her protagonists were using. Absent that, the book is often literally superficial.

Still, the book has a lot of upside -- but given that there is not exactly a huge selection of books covering this ground (unlike, say, quantum mechanics) to have this as nearly the sole choice is disappointing.

The book consists of short biographical essays of chemists who changed the world with the discoveries they made. These are very interesting stories that range from the 19th century through to the 21 century. Very well written and leads the reader to want to learn more about the scientists featured in the book.½

NYT