Thursday, August 29, 2002
Yet Another Bad Statistical Analysis by a Journalist

Here's a less than useful article on traffic accidents and fatalities along the interstate highways in Volusia County from today's Daytona Beach News-Journal. Featured inanities:
  • Quotes from a slow-driving tow-truck driver. (Hasn't the author or the driver in question considered that vehicles going slower than the traffic flow constitute some measure of a traffic hazard, too?).
  • A visit to the body shop where bloody wrecked cars are impounded.
  • This sequence of one-sentence paragraphs:
    State transportation officials evaluate a highway's relative safety with a complicated formula that takes into account the number of crashes, vehicles that use a roadway[,] and mileage driven.

    Applying that formula to Volusia's interstates shows there is an accident along I-95 for every two million miles driven.

    By contrast, there are about 1 accidents along I-4 for every two million miles driven.
    Okay. An accident in contrast to one accident. Unfortunately, while one wants to believe this is an editing mistake, there's not enough evidence from the rest of the piece to believe that the reporter understands numerical reality to a degree that he wouldn't say that one and one are in contrast to each other.

    Update: The article did include a typo: It should've read that the accident rate on I-4 is 1.5 accidents per two million miles.
  • What should be the money graph is next:
    Fatal accidents, meanwhile, occur on I-95 at a rate of one every 60 million miles driven. They're even less common on I-4, where the rate is one every 100 million miles. By comparison, the sun is 93 million miles from the earth.
    First, the rate of a fatal accident on I-95 is in comparison to one on I-4 in the same way as 1/6 is to 1/10. That's about a factor of 1.6 more frequent. So, we're not talking even twice as likely for a fatal accident on I-95 as on I-4. Does that justify the use elsewhere in the article of "far fewer" of the I-4 accidents leading to fatalities? (Some of this may have to do with the relative rates of actual accidents, which as noted above, were reported to be the same, but with some evidence that that's a screw up.)

    And there's that throwaway about the distance to the sun. 93 million miles? Really? Do I take I-4 or I-95 to get there?
  • There's no discussion of whether the differences in rates of accidents or fatalities could be accounted for by chance. The fact that there is a difference between the rate of fatal accidents may have something to do with speed, as the author is suggesting, or it could just be random. There are statistical tools that let one quantify the likelihood that chance accounts for the resuts. There's no evidence that the author is aware of such tools.
The article's bottom line tries to be that speed kills, but there's no convincing evidence presented in the article to that effect. It's just a collection of seemingly random statistics that don't make a convincing argument for anything.

The News-Journal, it should be noted, rarely met an accident on I-95 or I-4 that it didn't want to put a picture of on its front page or on its web site. Someone at the paper seems to have an almost fetishistic interest in crash scenes. It's kinda creepy.