Wednesday, April 12, 2006

...or just a comedy?

In our post below, http://bovinetb.blogspot.com/2006/04/comedy-of-errors.html , we pointed out a couple of extraordinary gaffs which were lurking in a recent paper submitted to 'Nature' the headline of which - if nothing else - is being widely quoted. A comment on the post reminded us that "badgers had been factored into the predictions" and for that at least we should be grateful. Or should we?

The Veterinary Times (April 10th) prints a letter from which we quote:

"When you are writing a scientific paper, the title is most important because it is all that most people will read".

So what do we make of the flyer on the cover of 'Nature' which asserts: " Bovine tuberculosis - Cattle Movements spread Disease" ?

The author the letter has not picked up on the sentence which we printed below re. the start date of BCMS mandatory registrations, but he has explored much more of the badgery data. And very interesting (weak?) it is too.

He says: "The paper is in fact a 6 page letter that reports on the statistical comparison of cattle movements against a number of other possible 'risk' factors described as "environmental, demographic, agricultural and climatic parameters".

Among these other risk factors, I could find no trace of any really 'fruity' predictors, like the observed presence of badgers in the farmyard, dead badgers in grazing fields, maize crop fields sited so that badgers were crossing grazing fields in early autumn, or even operations likely to disturb badgers, such as a national programme of tree felling on railway lines. Indeed apart from cattle movements, I could find no clear definition of any risk factor."

"The letter is surprisingly coy about specifying what the risk factors it compared were. "

The author explains that the information he required, is not available with the original 'letter' but as a supplementary addendum. And yes, as a 'predicter', badgers were mentioned, but not diseased badgers, and not on a comparable time scale to the BCMS data.

"The considerable datbase, built up by the SVS (State Veterinary Service) during the two decades prior to 1998, of locations where diseased badgers had been found was not mentioned; all I could find was the following remarkable statement:

"Badgers. Detailed information about badger distribution in Great Britain has not been possible to obtain".

" Information published in summary form for 1988 and 1997 was unavailable because the original data was collected by volunteers on the strict understanding that it remain confidential. Alternative, freely available data, are at best, patchy. The geographically most complete source from the Countryside Information System contains 1km resolution information on badger distribution from the Mammal Society, the Biological Records Centre and the British Deer Society surveys between 1965 and 1990. Even when aggregated to 10km. these records are unlikely to provide a very realistic representation of actual distribution" (Too goddam right - Amateur 'guessimates' from 1965 compared with BCMS data 2000 - 2003?? And they call this science? - you couldn't make it up. - ed)

The letter's author is more polite, "In other words, the only badger data entered into the analysis were of dubious value and covered a period between 1965 and 1990. By contrast the Cattle Movement data used in the statistical analyses covered the years 2000 - 2003"

"The elegant statistical analysis therefore compared worthless and irrelevant data from a 25 year period with extremely detailed and reliable data from a 4 year period - a decade later."

Describing the paper's authors as engaging in 'statistical gymnastics', the author also points out that even when a purchased animal proves to be the only reactor in a herd, this is still classed as a 'herd breakdown', as there is no other box to tick on SVS forms. "This is misleading. In any meaningful use of the term (breakdown) the disclosure of a purchased animal as a reactor, when no other animals in the herd react, either at the disclosing test or subsequently, is not a herd breakdown. "

He concludes "I am no statistician, but I am confident in asserting that no valid conclusions can be drawn from this 'statistical excercise' as to the relevant significance of cattle movements and the uncontrolled disease in a protected badger population".

Andrew J. Proud. BVSc, DVSM, MRCVS

We agree.

No comments: