1 of 2 people found the following review helpful
Clear and lively,
Verified Purchase(What is this?)
This review is from: Naked Statistics: Stripping the Dread from the Data (Hardcover)
In the wrong hands, statistics can be a dangerous thing. In the right hands, there's still no guarantee that the statistics you're reading in your newspaper or on your screen are much more trustworthy. In Naked Statistics, Charles Wheelan provides a little bit of a toolkit to give the uninitiated a fighting chance of at least recognising the potential pitfalls in the analysis they're reading and to ask some smart questions about its foundations.
Wheelan lays out his stall pretty early, making it clear that he's not a numbers for their own sake person. Numbers to him are only of any interest if there's a point, and he proceeds to explain why statistics are, in the final analysis, potentially a useful and friendly set of devices for explanation and, quite importantly, identification of potentially the best way to respond to a problem or need in government or business. In generally clear language he explains the basics of descriptive statistics, correlation, probability, polling and regression analysis, as well as outlining some common pitfalls and how they may be circumvented.
The writing style is clear and lively, and the examples he uses are practical and engaging, although the sports examples are very much oriented to the American reader, with abundant reference to Lebron James, quarterback ratings and at bat averages. In some ways, from that point of view it's something of a complement to Scorecasting, a book that applies principles of behavioural economics to sports (explaining, amongst other things, the reason for "Fergie time" and home field advantage).
But it isn't by any means all sports. One of the examples he dwells on is the use of Value at Risk (VaR) in banks, one of the infallible tools that told bankers that subprime mortgages would not pose a problem to their institutions, let alone precipitate a global financial crisis that would persist for six years and counting. The BBC documentary series Bankers focused on VaR in one programme, showing how excessive faith in the model had brought down MF Global, ably abetted by the hubris of former Goldman Sachs CEO and New Jersey Governor Jon Corzine, and how selective use of the model led to the infamous "London whale" debacle at JPMorgan.
Ironically, VaR was developed at JPMorgan, and one of the "quants" responsible, Till Guldimann, was interviewed for the BBC series, opining that VaR was good for measuring risk, but not for managing it. Wheelan reveals that actually it wasn't even much good for measurement, given that the data underlying the model was based on the boom period between the eighties and the turn of the millennium. He compares it to a faulty speedometer, worse than no speedometer at all, giving confidence where none is due.
Naturally there are a couple of issues.
His comparison between the Businessweek phenomenon, where business leaders featured on the cover for receiving high profile awards are guaranteed to be about to fail big time, and the Sports Illustrated effect, where athletes on a winning streak featured on the cover are about to hit a dry spell, one due to overconfidence, the other possibly because the winning streak was just a matter of luck, is valid enough. But he forgets to mention that sometimes behind the athlete's winning streak is a lifetime of practice, and the fall often due to injury (let's hear it for Carson Palmer, or in English Michael Owen).
Though mostly explanations are clear, I'm still not sure his distinction between "precision" and "accuracy" work that well, certainly for me.
More importantly, in one instance (p198 in the hardback) the term "dependent variable" appears where "independent variable" is intended, and he doesn't explain very well why it's the 99th percentile that represents the top 1%, not the 100th percentile (a little chart might have helped there, perhaps).
Those aside, however, I definitely found myself wishing I'd read this book before I studied econometrics. I think the veil would have lifted much earlier.