Sunday, July 06, 2014

Figures can Lie and Liars can figure....

A swift response to the recent Cull the Cows - [link] computer generated guff, came on Thursday from farming Minister, George Eustice. Speaking at he Livestock Event, and interviewed for BBC's Farming Today - [link] programme, the Minister was pretty scathing about the modeler's efforts.

Co author, Dr. Ellen Brooks-Pollock of Cambridge University described how the team had entered data into their model of potential TB bacterial transmission of:
Cattle to cattle infection, Cattle infections into to the environment and Cattle movements.
Which is pretty amazing when you realise that up to 52 per cent of badgers in areas of endemic zTuberculosis are carrying this disease. But we digress....

 Dr. Brooks-Pollock went on to explain that her work centred around 'idealised control measures'. Not if you're a cow about going to be shot on the back of a University's half baked theory it isn't: that's not 'idealised' at all. It's terminal.

But George Eustice was having none of it. He pointed out that despite epidemiological evidence to the contrary, the model had ignored badgers completely, focusing on cattle and making many false assumptions. Including one that the disease dies out naturally in the environment in 34 days.
For the pedants, badgers can sustain this disease, maintain body weight, breed and rear cubs for up to 8 years. Reactor cattle are shot - which Dr. Brooks-Pollock had in fact noticed, pointing out that 90 per cent of 'infected farms' posed no risk whatsoever. Quite correct. They are locked down, facing daily [in some cases] visits from infected wildlife and tests every 60 days with anything reacting to such 'visits' slaughtered.

Mr. Eustice stressed that computer models were only as good as the assumptions entered into them, and if those assumptions were wrong (as in this case) then the results were flawed. He also pointed out that a team of experienced Defra vets, and the Chief Scientist - [link] were exposing these flaws and that although the aim of this model was laudable, its assumptions were wrong.

 We can't take credit for the title of this posting: it appeared here - [link] with another succinct quote:
 'correlation does not imply causation'. 

No comments: