August 26, 2011

Suicides really have been lower in ChCh

News stories about monthly counts of road deaths, suicides, or other relatively rare events tend to cause statisticians to grind their teeth and mutter “Poisson variation”.  If you have two deaths a week  apart, some of the time they will fall in the same month and some of the time in different months, pretty much at random.  This makes the monthly totals very variable: if Christchurch averages about 7 suicides per month and there was nothing making this vary over time, you would expect in most years to see a month with as many as 11 suicides and another with as few as three.  That sort of variation is unavoidable, and doesn’t indicate that there is anything to explain.  It’s called “Poisson variation” because the “nothing to see here, move along” distribution for counts was investigated in the 19th century by French mathematician Simeon Denis Poisson, in a study of court judgments.

With only Poisson variation, a month with just one suicide would be very unusual, though. Only one month in 12 years would we be that fortunate if nothing but chance were operating.  The NZ Herald is quite right that suicide rates have been down in Christchurch — there is something to explain, and the explanation is reasonable. The Dominion Post does even better, giving multiple possible explanations for the dip.

The papers do lose points for not linking to the actual numbers released by the Chief Coroner, which I still haven’t been able to find.

 

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

    I’ve seen this kind of statistic before. In dire circumstances suicides drop. The theory is that when people see external reasons for their misery they tend not to self-blame.

    13 years ago

  • avatar
    Rachel Fewster

    Another thing to watch out for in November / December is whether there will be a spike in the birth sex ratio of girls to boys in Canterbury. Natural disasters seem often to be followed by a significantly higher proportion of baby girls some nine months later. A recent report is here:
    http://parenting.blogs.nytimes.com/2011/07/27/the-science-of-boys-and-girls/

    The only explanation I’ve seen for this is that boy-foetuses are more fragile than girl-foetuses, perhaps a ploy of Mother Nature to ensure that energy-hungry boys are more plentiful when times are good, whereas necessary-for-the-future-of-the-species girls are favoured when times are hard. It will be interesting to see whether this effect is evident in Canterbury.

    13 years ago

  • avatar
    Rachel Cunliffe

    The report is now up online here.

    13 years ago