May 3, 2012

Road deaths down

A record low in road deaths last month has been accompanied by an unusually good Herald story.   Andy Knackstedt from the Transport Agency is quoted as saying

“It’s too early to say what may or may not be responsible for the lower deaths over the course of one month.

“But we do know that over the long term, people are driving at speeds that are more appropriate to the conditions, that they’re looking to buy themselves and their families safer vehicles, that the engineers who design the roads are certainly making a big effort to make roads and roadsides more forgiving so if a crash does take place it doesn’t necessarily cost someone their life.”

That’s a good summary.  Luck certainly plays a role in the month-to-month variation, and these tend to be over-interpreted,  but the recent trends in road deaths are real — much stronger than could result from random variation.   We don’t know how much the state of the economy makes a difference, or more-careful driving following the rule changes, or many other possible explanations.

[Update: when I wrote this, I didn’t realise it was the top front-page story in the print edition, which  is definitely going beyond the limits of the data]

avatar

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »

Comments

  • avatar
    Mark B

    Yes, but…

    The road toll last year was 284 or 24/month. This story was based around the statistic that April had the lowest ever road toll of 12 deaths, or 12 less than the average/month for last year. In January, the toll was 36, 12 more than the average/month for last year. No story was generated.

    While there is a downward trend in road deaths to be celebrated and explained, this “record low” has been greatly over-interpreted, as you hint at.

    A worthy nomination for stat of the week.

    12 years ago

    • avatar
      Thomas Lumley

      To be fair, with Poisson(24) data, 36 or higher is more than twice as likely as 12 or less (0.0055 vs 0.013), and the decreasing trend has been visible for well over a year.

      The fact that it’s the record is perhaps over-hyped, but its a venial sin compared to traditional road-toll stories.

      12 years ago

      • avatar
        Mark B

        Really? If there is a downward trend in the road toll, and you have two outlying results (to a similar degree), the higher outlier is the more likely? That seems completely counterintuitive, and another reason why I am not a statistician.

        Looking at the data from the last 4y, there are 7 months for which the toll is 35% less than the average/month for the previous calendar year, and 4 months for which the toll is 35% more. That doesn’t seem to fit with your analysis.

        I still think the focus on the “record low” monthly level which is just an extreme outlier due to chance is unfortunate, regardless of the worse crimes committed in other stories. But maybe that is what is required to get the issue into the news?

        12 years ago

        • avatar
          Mark B

          Oops- should be at least 35% for both situations.

          12 years ago

        • avatar
          Thomas Lumley

          It’s not because of the trend, it’s because of the shape of the distribution, which is skewed positively.

          The square root of the number of deaths has a more nearly symmetric distribution.

          12 years ago

        • avatar

          Isn’t that a ‘feature’ of the distribution being lower bound by zero (i.e. there can not be less than zero accidents)?

          12 years ago

  • avatar
    Thomas Lumley

    I hadn’t realized that this was the top front-page headline in the dead-tree version of the paper.

    That’s definitely going too far for a one-month data point.

    12 years ago

  • avatar
    Thomas Lumley

    Luis,

    In some sense it’s a feature of the lower bound, but it is possible to come up with variables that are bounded below and unskewed or even negatively skewed.

    12 years ago