The Signal and the Noise: Why So Many Predictions Fail – but Some Don't is a 2012 book by Nate Silver detailing the art of using probability and statistics as applied to real-world circumstances. The book includes case studies from baseball, elections, climate change, the 2008 financial crisis, poker and weather forecasting. The book was the recipient of the 2013 Phi Beta Kappa Society book award in science. It has also been translated into several languages.
Synopsis The book emphasizes Silver's skill, which is the practical art of mathematical model building using probability and statistics. Silver takes a big-picture approach to using statistical tools, combining sources of unique data (e.g., timing a minor league ball player's fastball using a radar gun), with historical data and principles of sound statistical analysis, many of which are violated by many pollsters and pundits who nonetheless have important media roles. The book includes richly detailed case studies from baseball, elections, climate change, the 2008 financial crisis, poker, and weather forecasting. These different topics illustrate different statistical principles. For example, weather forecasting is used to introduce the idea of "calibration," or how well weather forecasts fit actual weather outcomes. There is much on the need for improved expressions of uncertainty in all statistical statements, reflecting ranges of probable outcomes and not just single "point estimates" like averages. Silver would like to see the media move away from vague terminology like "Obama has an edge in Ohio" or "Florida still a toss-up state" to probability statements, like "the probability of Obama winning the electoral college is 83%, while the expected fraction won by him of the popular vote is now 50.1% with an error range of ±2%". Such statements give odds on outcomes, including a 17% chance of Romney winning the electoral college. The shares of the popular vote similarly are ranges including outcomes in which Romney gets the most votes. What is highly probable is that the voting shares are in these ranges, but not whose share is highest; that's another probability question with closer odds. From such information, it's up to the consumer of such statements to use that information as best they can in dealing with an uncertain future in an age of information overload. That last idea frames Silver's entire narrative and motivates his pedagogical mission. Silver rejects much ideology taught with statistical method in colleges and universities today, specifically the "frequentist" approach of Ronald Fisher, originator of many classical statistical tests and methods. The problem Silver finds is a belief in perfect experimental, survey, or other designs, when data often comes from a variety of sources and idealized modeling assumptions rarely hold true. Often such models reduce complex questions to overly simple "hypothesis tests" using arbitrary "significance levels" to "accept or reject" a single parameter value. In contrast, the practical statistician first needs a sound understanding of how baseball, poker, elections or other uncertain processes work, what measures are reliable and which not, what scales of aggregation are useful, and then to utilize the statistical tool kit as well as possible. Silver believes in the need for extensive data sets, preferably collected over long periods of time, from which one can then use statistical techniques to incrementally change probabilities up or down relative to prior data. This Bayesian approach is named for the 18th-century minister Thomas Bayes who is credited for first deriving a simple formula for updating probabilities using new data. For Silver, the well-known method needs revitalizing as a broader paradigm for thinking about uncertainty, founded on learning and understanding gained incrementally, rather than through any single set of observations or an ideal model summarized by just a few key parameters. Part of that learning is the informal process of changing assumptions or the modeling approach, in the spirit of a craft whose goal is to devise the best betting odds on well-defined future events and their outcomes.
Release and sales Published in the United States on September 27, 2012, The Signal and The Noise reached the New York Times Best Sellers list as No. 12 for non-fiction hardback books after its first week in print. It dropped to No. 20 in the second week, before rising to No. 13 in the third, and remaining on the non-fiction hardback top 15 list for the following thirteen weeks, with a highest weekly ranking of No. 4. The book's already strong sales soared right after election night, November 6, jumping 800% and becoming the second best seller on Amazon.com. The Signal and the Noise (print edition) was named Amazon's No. 1 Best NonFiction Book for 2012. It was named by the Wall Street Journal as one of the ten best books of nonfiction published in 2012.
… excerpt ends here. Continue reading the full article.
