Bringing Sexy Back to Data

Nate Silver has certainly given the mundane topic of data a spanking new makeover with his book, The Signal and The Noise. Silver, who correctly predicted the winner in 49 out of 50 states during the 2008 presidential elections, said he was able to make a dent in political forecasts only because other political pundits were so bad at it.  Then he pointed to other industries such as finance and weather forecasting where predictions fail. It’s not all hopeless, however.

Silver has made data fun and buzz-worthy. He has recently infused data into such topics as finding the best burrito in America. The process to determine the 64 burrito-selling establishments nationwide that would be in the bracket involved something called a VORB, or Value Over Replacement Burrito, score, according to his blog. The judging system was based on 20 possible points in five categories: tortilla, main protein, other ingredients, presentation and overall flavor profile. The results led to a flurry of reactions (not all positive) on social media.

In my last post, I outlined three points Silver made in the book about why some predictions fail.  Silver goes onto make the following three suggestions on how to improve predictions:

  • Think probabilistically: Japanese authorities had built the nuclear plant to withstand a 8.6 earthquake in part because some seismologists concluded that anything larger was impossible. In March 2011, Japan was hit with a magnitude 9.1 earthquake. While Silver admits it is easy to say this after the fact, he writes that Japanese authorities should have considered historical earthquakes in other parts of the world with similar geological makeup to predict the maximum range of magnitude.
  • Know where you’re coming from: we must be aware of our weak points and biases. In a talk about the book, Silver points out a case study cited in Facebook Chief Operating Officer Sheryl Sandberg’s book, Lean in, where Human Resources executives were given two identical resumes whose only difference was the first names connoting their gender. The study found that not only were the men biased against female candidates but those who said they were not biased were actually found to be more so.
  • Try and err: Silver states that the only way to get better at forecasts is to make a lot of them, over and over again. To be sure, you won’t be investing your life savings on “practice round” stock market predictions. But he often quotes Bayes’s Theorem that states we should constantly update our forecasts any time we are presented with new information.

What has been your experience with data?

Leave a Reply

Your email address will not be published. Required fields are marked *