the theory that would not die (2011) by sharon bertsch mcgrayne is about the 200 year history of bayes rule which was marked by obscurity, secrecy, derision & finally glory but not before dividing the statistics community for many years <bayesians vs. frequentists | the fissure likely still exists & will continue to exist, it’s a philosophical almost religious difference> / computational resources now being both magnitudes more powerful & cheap, bayesian statistics is considered by some as the analytic tool of the 21st century
this is the most fascinating & engaging book i have read so far this year! / not only because we use bayes theory in our work but also discovering its breadth of applications such as: cracking the enigma code which helped the allies win the war | searching for the debris of air france flight 447 & navy searches <submarines, downed weapon-carrying b52s> | setting compensation insurance | more accurate probabilities of rare events which led to increased safety measures <warned but not heeded: three mile island, challenger space shuttle> | predicting election results, the first being kennedy vs. nixon <until the media resorted to exit polls | though nate silver famously predicted correctly state outcomes in 2008 using hierarchical bayes> | identifying the linkage between lung cancer & cigarette smoking | identifying the risk factors for cardiovascular disease <both with far reaching public health implications> | powering the google search engine | microsoft windows os & spam filtering | helping stanley the driverless car win the darpa challenge in 2005 <as well as driving google’s driverless car> | recommendation engines for movies, books, etc. | uncovering the authorship of 12 federalist papers
importantly too, many analytic techniques were developed to facilitate the application of bayes theory, among them: asymptotic approximations \ sequential analysis / gibbs sampling \ markov chain monte carlo <MCMC> simulation \ hidden markov models / kalman filter <though kalman denies that bayes rule had anything to do with his invention, it was proven mathematically by aoki that it can be derived from bayes rule>
at the very basic level, bayes rule is intuitive | one might even say, this is how we naturally think & make decisions: “initial beliefs + recent objective data = a new & improved belief” / after reading the drama surrounding bayes theory, it is alluring to declare sole allegiance to bayes \ i resist that because a good analyst will use the tool that is most appropriate to the problem at hand / knowing how to apply both bayesian & classical statistical techniques is the best position to take \ i highly recommend this book to anyone involved in data analysis | and to those who are not, enjoy the awesomeness of probability & statistics!
i learned about this book in the ny times
related current events: bayes theory will not work in the case of the missing malaysian plane because the prior is incorrect <an oil slick that was incorrectly thought to have been caused by the plane>