Anxiety Attack in Northampton; Lying with Statistics

Anxiety Attack in Northampton:
Nelson's telescope

Report from Radioactive Times Vol. 2 No. 2 September 1997

In September 1996, shortly after the Greenham bomb story became headline news, the British Medical Journal published an article by Sir Kenneth "BSE" Calman entitled Cancer: Science and Society and the Communication of Risk (1) A photograph of a lightning strike set the stage for an analysis of risk from carcinogens.

The following month, an article by Professor Paul Elliott and colleagues attempted to show that leukaemia clusters did not exist. Or more accurately, that we could use statistical sophistry to ignore them (2)

For example, in the case of the most famous cluster, the Sellafield leukaemia cluster, Elliot et al. state that since the TV producer who was originally looking at cancer in the workforce discovered that there was an excess of childhood leukaemia there only after he began to look, we should discount this, since it was not an initial hypothesis. The problem is called the Texas Sharpshooter. The extraordinary suggestion is that it is invalid to look for significance after a chance discovery of cluster. One has, apparently, to have a hypothesis first.

This specious attempt to deny the existence of leukaemia and cancer clusters was applied on 18 June 1997 by the BBC programme Anxiety Attack to the Northampton Pembroke Road leukaemia cluster. This cluster of five children with leukaemia in one road has been the focus of media interest in Northampton for some years and is touched on in Wings of Death in relation to the movement of radioactive fuel into railway sidings which originally ran along the back of Pembroke Road.

In their analysis, the BBC programme set out to contrast the ignorant parents (who believed that there might be some common cause of their children's deaths) with the analysis of the clever scientists, (who maintained that the cluster was entirely due to chance).

The Northampton Health Authority's own report, which the BBC programme drew on heavily, began with the fact that any truly random dispersion of points on a map would naturally have some points which were close together, or clustered.
Drawing a circle around these after the event would be like the Texas sharpshooter.

In the report, they reproduce a 10 x 10 grid in which the distribution of 40 random numbers are supposed to show a cluster. This diagram is reproduced here (Fig. 1)and was shown on TV.

,b> Bogus statistics

Our attempt to reproduce the effect failed. This is because such a cluster is actually very unlikely, and not a simple chance occurrence.
Not only that, but since the number of cases of leukaemia they were considering in the whole area was 28 and not 40, they should have used 28 points.

Calculation using Poisson statistics enables us to show that in the 28-event case the probability is 0.016 or one chance in 63 of any of the one hundred squares containing four points.

But if this were not enough, the TV programme did not even show the real leukaemia map (Fig. 2) Nor did editions of the Northampton report sent out to at least two people we spoke with.


References

1 BMJ 313, 799-802

2 BMJ 313, 863-6


If you are seeing this page full screen (i.e. without a navigation bar on the left) you can't see how the rest of the site is organised.
This Home page link takes you to the index page, which has links to all the topics we discuss on the site [only use it if this page is full screen]
Use the Flawed Epidemiology button to see what else we have to say on this topic.

This page was last updated May 2001