Fw: Valentine's Day on sp!ked (serious statistical strangulation)

Daniel Davies d_squared_2002 at yahoo.co.uk
Wed Feb 14 06:13:27 PST 2001


Warning: below follows an extended rant about statistical misuse. It's completely over the top, but I'm really rather sick of people maligning reputable and hard-working professionals who work with the data every day, on the basis of a journalist's half-thought-out screed.

--- Russell Grinker <grinker at mweb.co.za> wrote:


> From the people who brought you LM...

And haven't learned much about statistics since those days, I see. Perhaps, Russell, you might have a quiet word with Josie Appleton that if she's going to start slagging off the underappreciated and extremely hard-wroking professionals who compile our national statistics, she might want to run her piece past someone who's got a clue what they're talking about first. I'd almost volunteer for the job myself if I thought it would prevent train wrecks like the one described below, entitled "Statistical Abuse", http://www.spiked-online.com/Articles/0000000054A4.htm

....

The key paragraph, on which the whole rest of the article depends, runs as follows:

---------------------------

"If a woman in the UK suffers domestic violence every six to 20 seconds, between 1.5million and 5.25million incidents of domestic violence occur every year. Apart from the size of the margin - a difference of four million violent incidents is quite a lot - something about this seems at odds with everyday experience.

The UK has a female population of 30million; 23.5 million are over 18. The number of women cohabiting with a man is likely to be substantially lower than these figures. If a woman in the UK suffers domestic violence every six to 20 seconds, then, up to one third of cohabiting women are abused each year. Or to put it another way, one in every three or four of your female friends and colleagues could be suffering from domestic abuse. Is that what you find, in your life? "

----------------------

This is a genuine howler. In order to go from "5.25 million incidents" to "one in three women", you have to make the assumption that every single act of domestic violence happens to a new individual -- that each incident increments the abused population by one.

How likely is it that a woman will suffer precisely one act of domestic abuse in a given year? The assertion that "every six seconds, a woman is abused by her husband -- and she's getting mighty sick of it" is a tasteless joke; but Spiked seems to believe something equally as ridiculous.

A lesser, but still irritating complaint is that Josie suffers the occupational hazard of associating with the LM crowd - the belief that there is something sinister and important about confidence intervals. "Apart from the size of the margin - a difference of four million violent incidents is quite a lot " -- yes it is, but it's exactly the same as the difference between six seconds and twenty seconds, which is the whole point.

Further on down the page, we have:

"In the claim that more than one in four women have been subject to domestic violence, Jayne Mooney, widely credited as an authority on the scale of domestic violence, took being forced to do menial and trivial tasks as an example of domestic abuse. It is odd, really, that the figure she came up with was not higher."

Which comes very close to implying that the earlier calculation was made on the basis of anything other than the official crime statistics, which would be a lie.

Another case in which the article comes even closer to actually lying is in the grossly misleading citation halfway down the page:

" - More than one in four women have been subject to domestic violence (4). "

(4) in the footnotes is Jayne Mooney's study, the loony loony one which, we are told, counted being asked to do the washing up as "domestic abuse". The actual source for this figure, however, which is helpfully linked to at the foot of the article, is C. Mirrlees-Black (1999) Domestic Violence: Findings from a new British Crime Survey self-completion questionnaire. Home Office Research Study 191, using data from the 1996 British Crime Survey.

Meanwhile, further still down this increasingly horrifying heap of pernicious bollocks, we have:

" One survey, published on netdoctor, claims that 13 percent of 15- to 24-year-old men have been assaulted by their partners because of pre-menstrual syndrome (7). Two in three men say their partners have suffered from PMS. Does this mean that one in five PMS sufferers has assaulted her partner? Was this while they themselves were being abused - or at a different time of the month? "

which is of course a species of the same howler made above, but with the added bonus of being based on "one survey, published on netdoctor" which, because it didn't publish a margin of error, is regarded as pinpoint.

The article concludes with "And I have to ask again, can this possibly be true? Is that what you find, in your life?"

Well, I don't believe that this sort of statistical butchery is true, but I'm afraid I find it in my life all too often.

===== “It is necessarily part of the business of a banker to maintain appearances and to profess a conventional respectability which is more than human. Life-long practices of this kind make them the most romantic and the least realistic of men” -- JM Keynes

____________________________________________________________ Do You Yahoo!? Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk or your free @yahoo.ie address at http://mail.yahoo.ie



More information about the lbo-talk mailing list