On February 7, 1978 I was sitting comfortably on the lanai in Honolulu Hawaii’s charming Manoa Valley. The morning edition of the Honolulu Star Bulletin was on my lap and every story above the fold was about the giant snow-storm that had completely paralyzed the northeastern United States; a storm that would be known ever-after as the “Blizzard of ’78”. This was long before the current fetish for giving a name to every weather system more serious than a light drizzle.
Despite having spent most of my life in New England, this once in a century snow-storm had struck while I was living in the tropics. If you are not from the northeast you should understand that surviving severe snow storms is an unofficial badge of honor among the natives. Being a New Englander and missing the Blizzard of ’78 (by being in Hawaii no less!) was akin to dereliction of duty. Your right to call yourself a New Englander could be revoked for less than that. Another year went by and I returned to the northeast; I overcame the shame and humiliation of missing the big snow storm.
Where Were You During the Big Storm?
Fast-forward 34 years. Two weeks ago we were making final preparations for a vacation trip to Saint Croix (in the US Virgin Islands). Weather reports began to appear about a snow-storm that might begin on Thursday February 7th. We were supposed leave early on Wednesday the 6th, so the night before leaving we asked a neighbor if he would clear the driveway if more than a few inches of snow fell; he said he’d take care of it. On the way to Bradley Airport to catch our plane, we heard forecasts calling for as much as 6-10 inches of accumulation. Not long after landing in Saint Croix, the forecasts were beginning to describe a potentially historic snowfall.
As the second great nor’easter of my lifetime bore down on coastal New England, all I could do was shift my eyes from the swaying palm trees on the beach in Saint Croix to the iPhone in my hand to check the New England snow accumulation update from home. Our neighbor texted to let us know that 30 inches had fallen (and had been removed) from the driveway; he included a photo. So not only did I miss the Blizzard of ’78, but also I missed storm Nemo (don’t you think “the Blizzard of 2013” would have been a catchier name?). What was the probability of me missing both of these monster storms?
Silver in the Suitcase
Nate Silver is an author who writes on the topics of prediction, forecasting and probability. During the last presidential election reading his FiveThirtyEight blog (in the New York Times) was on my daily must-do list. Silver is an entertaining writer, but his real talent is in squeezing mountains of data to come up with clear intelligent explanations of the data’s meaning; he is particularly effective at interpreting and explaining political polling results. His analyses are so compelling because his treatment of data is completely exhaustive.
Some time before the November election I purchased a copy of Silver’s new book “The Signal and the Noise – Why so Many Predictions Fail, but Some Don’t”. It and one other half-started book came with me to Saint Croix; sadly the other book remains unfinished, but after picking it up I could not put The Signal and the Noise down. DISCLAIMER: the prose is not brilliant and the overall flow of the book is a little disjointed, IMHO. However, Silver’s management of and statistical insights into data sets are remarkable.
Risk Assessment = Risk Prediction
Silver’s work is interesting to me in part because in Massachusetts we have a “risk-based” waste site cleanup program. This means that cleanups are designed to reduce health and environmental risks to a level determined to be generally acceptable. This determination of acceptability is based upon a prediction of the risk posed by a waste site; these predictions are also known as “site specific risk assessments” or, in Massachusetts, “site specific risk characterizations”.
As Silver points out many times in The Signal and the Noise, to have value the accuracy of a prediction needs to be tested against future events; does the prediction forecast future events with reasonable accuracy? If so, then it had value, if not, then the prediction had little or possibly even negative value. Negative value can occur when people take unnecessary, possibly dangerous and/or costly countermeasures in response to a prediction.
By the time I finished the book I had to wonder what would Silver have to say about the probability of me being in the tropics for both the Blizzard of ’78 and storm Nemo? Most likely he’d say my sample size was too small for predictive purposes. My hope is that he’d add that I should continue going to the tropics from time to time while simultaneously tracking big snow storm occurrences. You have to have a lot of data for this type of work.