Never Mind the Science, Here’s the Bollocks
by Edmund Harbord

All the men on the tube had their legs crossed this morning. The reason was on the front page of Metro: “Electric blankets make you sterile”. This was a particularly harrowing example of the Friday Morning Health Scare, but there have been outbreaks all over the media. Broadsheets and tabloids alike have been disfigured by splotches: Aspirin causes cancer! Aspirin prevents heart disease! MMR causes autism!

“People lie. They lie to surveyors and they lie to themselves. 15% of teenage boys claim to have heard about “gonaditus”, a fictious sexually transmitted disease”

Speaking of bollocks, why not generate your very own health scare? First, we need to find out what the people are doing and tell them to stop. Great surveys should be conducted into peoples lifestyles.

Here lies the problem: people lie. They lie to surveyors and they lie to themselves. 15% of teenage boys claim to have heard about “gonaditus”, a fictious sexually transmitted disease, while opinion polls asking how people voted in 1992 reveal a Labour landslide undetected by the general election.

They also “misreport”, which is like lying only more decorative. Do you know how much fat you ate last week? The week before? The second week in February? Do you eat the same amount each week? How many units did you drink? Are you going to tell the pretty nurse you don’t know? Or make up something that “sounds right”?

All this inaccurate data is subsequently fed into the computer for a serious pounding. Low and behold, there are correlations between some suitably politically incorrect habit and some nasty lurgy. (“Passive smoking causes cancer” “Racism makes you blind”) Unfortunately for the disease detectives, correlation does not mean caused by. There could be a third, unknown factor: houses with refrigerators are more likely to burgled. Rather than thieves being drawn to friges, such households may be more affluent and make more profitable targets.

We could be confusing cause with effect: criminologist William Nagel  noted in the 70’s that societies with higher crime rates had higher rates of imprisonment, concluding that reducing rate of imprisonment would reduce the rate of crime. As locked up crims are in no position to commit crimes, Nagel must assume that imprisoning them makes the law abiding more criminal.

Statistics do not lie. Unless bad statistics were banged up. Then the others would lie.

“The Lancet study which linked MMR with autism was based on seven children. The breast cancer was based on 13. Studies of half a million children and 49,000 women  found there was no effect.”

Or it could be pure chance: 95% confidence, a common requirement for acceptance for publication, only means 1 in 20 times.  There is a strong correlation between the decline of piracy and global warming, although the climate pornographers of The Independent have yet to pick up on it.

Any number of lifestyle factors can be cross-referenced with the fashionable diseases du jour. We can stack the deck in our favour by looking at loads of naughtiness/illness combos: connection with  Alzheimer’s? Nah. Blindness? Nah. Cancer? If  the dice don’t come up in your favour, roll again: remember, one time in twenty your number will come up. At the Casino of Media Terror you can win big, and it’s everyone else who loses. With ten diseases and ten putative causes, we can generate 100 combinations and expect 5 correlations (Monday’s headline, Tuesday’s headline,..)

The Lancet study which linked MMR with autism was based on seven children. The breast cancer was based on 13. Studies of half a million children and 49,000 women found there was no effect. Small number investigations are just too small to be meaningful. Bonus points for using an odd number of trial subjects so there are always a majority that are healthier, making your random cause either a miracle cure or a wicked scourge. Either will do.

But suppose the big survey data show no evidence of a connection. This is no good: “passive eating is completely bloody irrelevant to cancer rates” will commission no editorials in The Times nor generate fawning segments on Richard and Judy. There may not be a correlation in the entire population, but do not despair: Time to play “Salami statistics”. Reduce the sample down to women. Chinese women. Pregnant Chinese women. Pregnant Chinese women aged 20-25. No? What about aged 21-27? Sooner or later, the sample will be small enough that you’ll generate a result purely by chance, and best of all you’ll quote the total number of subjects (50,000) not the number you didn’t exclude (3).

A smoker is 24 times more likely to develop lung cancer than a non-smoker. But we don’t require such irrefutable evidence. If a one- in – million chance becomes a two in a million chance: “100% INCREASE IN CANCER RATES”.

It’s time to write our press release: “one-in-10,000 risk (maybe)” won’t hold the presses, so lets say “50,000 TO DIE IN PASSIVE EATING MASSACRE”. That might elicit a more proactive response. Anyone with a calculator can play “Fantasy Death Toll”. Professor Richard Lacey was predicting 500,000 deaths from CJD. At the time of writing, there are 71 causalities. Bollocks to bullocks.

So it’s headline in the press: millions to die from a cause unheard of yesterday which will be forgotten by tomorrow. It all could have been avoided with a bit of common sense and a smattering of maths. Never mind the science, here are the bollocks.