Innumeracy

"One death is a tragedy. One million deaths is a statistic." --Stalin

Imagine you read about a study of a new cancer prevention regime. Doctors conclude that this regime cuts the odds of cancer in half. Would you do it?

Bear in mind, the 50% risk reduction could mean that two groups of 1,000 people were studied. One group underwent the treatment, the other (the control group) did not. In the control group, ten people eventually received cancer diagnoses, compared with five in the experimental group. Statistically significant? Probably. Extremely meaningful? Who knows?

Obviously, your answer to the original question--would you adopt the regime?--depends on a variety of factors. How onerous is the treatment? If relatively simple (drink a glass of orange juice every day), you do it. If relatively heinous (eat a brussel sprout), you don't. But either way, numbers lacking context--or clarifying information--serve a limited purpose.

As we follow the news about the continuing investigations into the BP oil spill, we think about statistics without context. Today we read that, last September, an audit found that the oil rig needed some 390 repairs, many of which remained uncompleted at the time of the disaster. 390 sounds like a large number, but is it? How does it compare to the average number of safety violations at other drilling platforms?

We imagine that, if OSHA came to look at the Solipsist's office, a roughly 100-square foot room occupied by one employee, they would find safety violations: papers everywhere, questionable ventilation, clutter--to say nothing of the gophers. How many more violations occur on a massive oil rig with over 100 employees and 18 gazillion moving parts?

Indeed, the sheer number of violations is less important than what the violations are. The news today spoke of an alarm system on the Deepwater Horizon that was frequently "inhibited" by the crew and thus not functioning at full capacity on the day of the explosion. This diminished capacity may have played a role in the deaths of eleven crew members, but the newspaper article also noted that workers were, indeed, allowed to inhibit the alarm system because they "didn't want people woke up at 3 a.m. from false alarms." In other words, this was a "permissible" violation--or, not a violation at all. (If anything, it screams for the need to improve the alarm mechanism so that it's not so prone to false alarms that it necessitates crew members shutting down the whole system.)

This is not meant as an apologia for Transocean or BP. Obviously, they did something wrong, and, even if they did everything right, their response systems have proven hopelessly inadequate. We write this simply as a request for the media to provide full context when presenting spectacular statistics.

(Image from The New York Times)