Wednesday, September 02, 2009

Lies, damned lies and statistics

A doctor writing in Online Opinion neatly illustrates a statistical trick used to spin messages, usually fooling journalists into parroting figures that mislead readers. The same trick is used by prohibitionists -- and also swallowed uncritically by lazy journalists.

Dr Andrew Gunn, editor of New Doctor, writes:
The science underlying modern pharmaceuticals gets distorted by commercial considerations. Manufacturers quite naturally seek to present their products favourably, for instance, by highlighting relative risk reduction when absolute risk reduction gives a more accurate picture of a drug’s efficacy. If trials suggest a medication will reduce a patient’s risk of stroke over the next five years from, say, 4 per cent to 3 per cent then the relative risk reduction is an impressive-sounding 25 per cent (the risk reduces by one fourth) but the absolute risk reduction is a measly 1 per cent (one chance in a hundred).
Prohibitionists reverse the effect, as noted elsewhere in this blog, typically in alarmist messages about cannabis and psychosis. A 40% increase in risk sounds horrifying but if it's happening among far less than 1% of users it falls into perspective.

They should teach this statistical trick in journalism school.


Anonymous said...

An interesting comment, but why not use real numbers.

The author says a 40% increase on less than 1% is not significant.

So lets say 1 million people using. Natural pyschosis 10,000.
Additional pyschosis 4,000 people.

Still a big number.

The Editor said...

That's very true. But, firstly, the link between cannabis and psychosis has still not shown to be causal (some research does, other research shows the opposite). Also psychosis hasn't significantly increased along with the massive increase in cannabis use in the west over the past four decades.

But to the statistical point, Anon's figures would still leave 999,996,000 healthy people being persecuted by ineffective laws that haven't stopped the problems anyway. I'm not saying cannabis is harmless, just that the harms get exaggerated and prohibition is not the way to solve problems.