July 3, 2017


  • The Australian has a story about religion in Australia, with the figure of “a touch under 70 per cent nominating a religious affiliation.” That’s only true if you include the 10.6% “Religious affiliation not stated” as nominating a religious affiliation.  The true figure is 56.8%.  You could stretch that to 63.5% of those who answered, but that’s assuming (at least implicitly) that the no-answers are just as likely to be religious as those who stated an affiliation. (via John Quiggin)
  • From the Dallas Morning News:Steve Doud, a subscriber from Plano, emailed me to say he’d read something in the June 21 Dallas Morning News that couldn’t possibly be true.An eight-paragraph Washington Post article on page 10A reported on a national study about kids and guns. The last sentence said 4.2 percent of American kids have witnessed a shooting in the past year.
  • There’s news in the UK claiming that Ryanair’s “random” seat allocation for people who don’t pay extra to pick a set isn’t random and favours middle seats, including something from Oxford University statisticians. It’s clear that the seats are not a simple random sample of all seats on the plane, but that isn’t what Ryanair is claiming. The airline’s defence, that the people who pay to pick seats usually pick the better ones, seems pretty plausible to me.
  • Buzzfeed says the US is going to spend $7.1 billion on food for the Fourth of July weekend. If you think about this as 321 million people for a weekend plus a holiday, it’s not as impressive.  Maybe they’re only counting some types of food?  This violates two of the three basic rules for big dollar numbers: they need to be per-capita, inflation-adjusted, and compared to something. (via Fred Clark)
  • “A sloppy attitude towards statistics has led to exaggerated and unjustified claims becoming commonplace in science, according to one of Britain’s most eminent statisticians.” David Spiegelhalter, reported in the Guardian

And finally, a picture of one of the causes of overdiagnosis in cancer — and one of the causes of illusory improvements in survival from earlier screening.


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 »


  • avatar
    megan pledger

    The 4.2% figure has obviously been mis-specified but how far away would the actual percentage be?

    How common an occurrence are street riots or bombs going off? And do they occur independently of people being shot?

    And what does “witnessing a shooting” mean? At first thought it means actually seeing someone doing the shooting. But what about the kids locked inside a classroom hearing kids and teachers being shot on the other side? Or not seeing the actual shot itself but seeing the result? The latter examples sounds like witnessing a shooting even if the children didn’t see the actual shots being fired.

    And what does “place in real life” mean? Obviously not in a video game or movie or tv programme but what about a documentary or tv news e.g. JFK’s assignation?

    And if “place” part of the clause excludes video/movie footage then, if the question was asked of adults, would the weapon’s operator on a drone, who only sees the resulting deaths via video footage, answer yes or no. And how is that different from seeing the same footage in the news or as part of a tv show e.g. The Good Wife, s03E09?

    It seems a really tough question to define well.

    3 months ago