Friday, September 21, 2007

The law of unintended consequences

Or as we say - Sod's law. As the barricades go up once again on livestock farms across the country, cattle farmers are counting the cost of preMT in spades.

When FMD (Foot and Mouth Disease) was discovered on a farm in Surrey, early August, all livestock movements came to a halt. The outbreak, traced back to the Pirbright site operated by Merial Laboratories and Defra's flagship Institute of Animal Health is well covered by the indefatigable warmwell , but as a consequence of the industry 'shut down', cattle pre-movement tested for sales during August have not moved.

When trading between farms resumed for the briefest of windows mid September, many of these preMT tests were 'time expired'; that is to say the selling opportunity was breached as the testing had been done more than 60 days ago. Some farmers tested again that week - if they were lucky enough to get a vet at very short notice - only to be stopped again by the discovery of healed lesions and FMD antibodies in livestock at Egham.

Auctioneers and farmers are telling us that thousands of head of cattle are stuck on farms, needing yet another preMT before sale. But now, as routine autumn testing starts in earnest, veterinary practises are several personel short, having loaned vets to Surrey. And they are telling clients that they cannot cope with extra preMT until at least November.

September is the traditional 'hand over' time when farms change owners or tenants, and this too, is now put on hold, due to snarlups in the trading system. Late October / November, which is Defra's new window for loosening up trading restrictions, (assuming no new FMD cases pop up) is the wrong time of the year for autumn suckled calf sellers to sell. And the delay in moving stock puts horrendous pressure on dairy farms, regularly trading calves and newly calved milking cattle. And it comes at a time when wheat, maize and other starch prices are rising at an extraordinary rate.

So a double whammy of an FMD 'leak' in Surrey, combined with the costly, bureaucratic and cumbersome 'comfort blanket' of preMT has led to severe overstocking and consequential feeding and possibly housing problems, on livestock farms who are on annual or two year bTb testing regimes.

The eyes of the media are on Surrey, but the law of unintended consequences, is alive and kicking on thousands of cattle farms from Staffordshire to the tip of Cornwall.

Sunday, September 16, 2007

Comments (and errors)

Some of us are computer literate, and some - are just a bit 'challenged', I think.

After being inundated with 'Anonymous' comments, reluctantly, we deleted that option - and apparently zapped all comments options. That was clever. So, Matt 5, read the book, and reinstalled 'Any Comments', but on scrolling down the blogger tool kit to verify the change, altered the next box to 'Allow No Comments' - at all. That was really clever. And not intentional.

About as good as the ISG's simple mathematical model which received an exclusive, professional brief peer review in our posting below.

Our modeler also challenged the ISG computer's maths on the incidence of New herd breakdowns, which were calculated as doubling 'every 4 and a half years'.

Incidence of New Breakdowns (7.21) on p148 - error on the first line
[The ISG quote] a compound IR of 15% - I’d be happy with that.
Compounding annually at 15% means it would double in 5 yrs not 4.5.
That’s out by 10%. A fairly big error straight off.


The maths: 1 x 1.15^5 = 2.01136

1 x 1.15^4.95 = 1.99735


Ten per cent is a lot to be adrift by - but hey, the ISG did describe their model as 'simple'.

Another 'error' was found by our eagle eyed contributers, in the recently published SE 3013 Pathogenosis project report. The stats on Gamma Interferon use within the trial we found odd, as was the conclusion of the mathematical calculations. Were some of the conclusions, including, possibly, this one, used to fuel a 'simple modeling exercise' too?

We quote from the point made to VLA on their figures for the IFN trial:
If 23 out of 96 tested negative to both of two tests, then logic dictates that at least 23 must have tested negative to either one of the two tests. So the figure of 9 in the last sentence quoted doesn't square with the earlier statement. But the higher figure of 30 reactors with unconfirmed bTB which tested positive could be correct. If so, 42% of reactors with positive IFN results but no confirmed bTB seems a large proportion.

And the answer confirmed our query:
"Thank you for identifying this error for which I am responsible: "23 out of 96 tested negative to both tests" should read " 6 out of 96 tested negative to both tests" . 23 out of 96 animals tested negative to either test".

So, 23 out of 96 (23.958 per cent) becomes 6 out of 96 (6.25 per cent)- an error of almst 18 per cent. Our modeler described the ISG's mathematical error of 10 per cent as 'fairly big". We can't begin to guess how he would react to a gaff of 17.708 per cent.

So when considering veterinary and epidemiological options based on these multi million £ research projects, readers would be wise to keep a firm handle on their brains and common sense, and if a conclusion looks odd, to question both the conclusion and the manner in which it was reached.

Thursday, September 13, 2007

"A simple mathematical model"

We are grateful for the over view of a mathematical modeler, with an MSc degree in the subject, self employed and freelance in his profession.
He has looked at the basis of the ISG data for us, and comments thus:

"One of the more telling sentences is the penultimate one of section 7.19: "

“The ISG chose to use a relatively simple model rather than a more complicated and potentially more realistic model because the objectives were to reach rather broad conclusions, and because the more realistic models require the specification of many essential unknown features”.


"In mathematical modelling terms, this basically means:"

a) It is a simple model – therefore not even close to modeling a ‘realistic’ scenario. Waste of time basically.

b) The objective was to reach a broad conclusion – ie nothing of substance again.

c) There were many essential unknown features – again the more realistic model that might be of some use can not be used as they don’t have the information.


Our modeler continues:

How can this model, using such basic assumptions and missing key information (because of unknown but essential features) be relied upon for real world decisions that affect people and livelihoods?! It is fine for an academic discussion but that is where it should be kept.


So, it would appear that by opting for a simplified, broad brush model into which to put generalised assumptions, the ISG has turned an embryonic mathematical 'science' into an art form which can be tweaked to achieve their intended answer We would add, that to see the eminent professor actually appear proud of his acceptance of such a political bridle in which to 'work' was quite extraordinary. Illuminating, but extraordinary.

But we digress: in modeling terms, our independent expert describes such input as ;

"simple squared equals stupid". Referring to the fact that the ISG have used a simple model to which they have applied simple assumptions.


We referred to such blind faith in the machinations of skewed input into computers in our post here and also drew our reader's attention to one such extremely broad 'hypothesis', on the relative importance offered to the sacred machine of 2 parts cattle v. 1 part badger, which extruded 50 per cent cattle to cattle transmission - and the inevitable cattle carnage. This input data could have been much more accurate, if information from the Tb 99 tomes had been used. Lord knows we spent long enough filled the goddamn things in.

We note that the ISG have recorded manure spreading as a factor encountered on the vast majority of breakdown farms. Yes. Cows do have that unfortunate habit, and the resultant organic manure is spread. And?

Sunday, September 09, 2007

Using the 'science'.

We have, and will continue to be highly critical of the politically skewed RBCT Badger Dispersal Trial. It told us little that we did not know ten years ago, except to reiterate how not to deal with a serious, zoonotic disease situation in a wildlife reservoir.

But the further one delves into the ISG 290 page final report, the more snippets emerge which, although not highlighted by the authors, do not support the unequivocal conclusions of its chairman. (that badger culling, as it was conducted in the 'trial', has no place in the control of bTb).

A comment from Jim on a posting below first drew our attention to the published results of the cull areas of the trial, but in much more detail;

I've done some "mathematical modelling" (based entirely on the data in the ISG Report), and this "suggests" (to use modellers' language) that, if you take a cull area of 800km2 (as opposed to the RBCT area of 100km2), then the net number of herd breakdowns saved (i.e. the number of breakdowns prevented in the cull area minus the number of breakdowns induced by the perturbation effect in the 2km ring outside the cull area) would be 121.9. This is to be compared with the RBCT figure of 1.4 net breakdowns saved per 100km2 (or 11.2 for eight separate areas of 100km2 each). This result is simply a (mathematical) function of the fact that the area culled is proportionately larger when compared to the perturbed area just outside the cull area.

and the really important bit, on which the ISG did not dwell;
I've also taken account of the fact that the ISG found that the beneficial effect of proactive culling was about 50% in an inner core area 4 - 5km from the trial boundary - see para. 5.15 and fig. 5.1B - though this doesn't seem to be a finding they are keen to highlight
.
So should the data that the centre of Bourne's cull areas managed to avoid the problem of his 'edges' - or any other part of the man - have been given more prominence? That depends from where the political steer was coming. Our commentator continues:
Surely, if the mathematical modellers are saying that something would have a beneficial effect (i.e. culling over a larger area), then it's up to the politicians to find the will and the means to achieve that (or not, as the case may be). Instead, no doubt they'll hide behind the good professor, who has assured them it isn't practicable - and in the meantime they've conveniently sacked all the people with the necessary expertise.

Jim stresses that he is not advocating much larger cull areas as the only or even the necessary way forward, preferring a more targetted approach, but he points out what can be achieved through the use of mathematical modelling.

"If you take the ISG's own figures at face value, the question for the politicians is why not devise a way to carry out a cull over larger areas, as the mathematics shows that this will have a significant beneficial effect? "For the avoidance of doubt", as they say, (a) in setting my calculator to work on the ISG data I'm not accepting that their report is anything other than fatally flawed, (b) if anyone thinks my maths is wonky, feel free to say so (though it appears to be in line with paras. 5.39 to 5.42 and fig. 5.4 of the ISG report), and (c) I realise that questions of badger ecology might influence the effects of a cull over a larger area (I'm just trying to show how a bit of number-crunching can affect one's conclusion

Another observation, this time from Dr. Nick Fenwick in his evidence to the Welsh Assembly makes the same point;

"There is a graph on page 99 of the ISG report in which the beneficial effects of what we would regard as not exactly efficient badger culling are plotted against the distance from the outside of the culling area. The closer you are to the outside of the culling area, the more significant the perturbation effect would be, because that is where the badgers are travelling in and out of the culling area. The badgers from outside the culling area are too far away to travel to the core of the culling area. So, ultimately, perturbation is not a problem in those core areas.

The relationship between the beneficial effect and the distance from the perimeter of the culling area is plotted. The average figure for the effect over five years is in the early twenties: there is around a 23 per cent or 24 per cent reduction in TB. However, in terms of the core areas, where we can ignore perturbation, the reduction in the incidence of TB is down to 50 per cent.

In addition to that, it is worth bearing in mind that, on that same page in the ISG report, there is also a plot of the effect after each year of culling. After the fourth year, the reduction in the incidence was some 32 per cent — I am estimating these figures from the graph. There was a 32 per cent beneficial effect, compared with around an 8 per cent negative effect on the periphery of the culling area. If you combine those two aspects, one might ask what the reduction in the incidence of TB in the core area was after the fourth cull, because that figure of 50 per cent is likely to be increased significantly.

So, in the fifth year of culling, you could imagine perhaps a 60 per cent or 70 per cent reduction in the incidence of TB, which would be a major breakthrough in TB control and would reverse the decline in our TB status that started in 1986."


We agree, but would suggest that annual culls are merely a way to prolong this agony. An intense targetted cull of weeks, would draw out dispersers from other social groups, thus leaving a 'clean ring' of badgers for probably several years. Do not forget that the length of time from exposure of a sentinel cow to a skin test flag up reaction from UK strains of bTb is 221 days. Just 7 months. (Not the seven years the strains take to manifest in Australia) There is absolutely no reason to suppose that the UK system of two 60 day skin tests plus a follow up six month check test for a herd, will not clear the problem, providing the source of infection is identified and removed.

And although they may not appreciate our pointing it out, the ISG final report has shown a huge reduction in cattle Tb at the centre of the cull zones, even and especially given their ineffective, irregular 8 night hit-and-run visits. And no cattle controls either, apart from regular skin tests. How odd, that they felt unable to share this.

Thursday, September 06, 2007

"This is the answer ....

... now, what was the question?"

The ISG final report from the RBCT Badger Dispersal Trial was categoric in its recommendations. Badger culling, in the way that it was conducted during their trial, has absolutely no place in the GB armoury against bTb - unless as he said in our posting here it was to act as an inducement for the cattle measures which Professor Bourne openly told veterinary colleagues at the start of the excercise, would be their only option, at the end of it.

We have explored the difference between the trial protocol as described by Professor, Sir John (now Lord?) Krebs here
and how it differed substantially and critically from what the ISG delivered in practise. And from the Welsh Assembly's examination of various industry reactions to the report's findings, we quote this gem from Mr. Rooney, speaking for the CLA:

" Perhaps I might preface my remarks by saying that I was brought up as a scientist; it was not in this discipline, but scientific principles hold, whatever the discipline. One of the things that I was taught was that, in designing an experiment to try to address an issue or a problem, you may not like the results, but you accept them. I find it deeply shocking that responsible scientists should have been prepared to undertake a research study having been told at the outset that there is a conclusion that they are not allowed to reach. I find that utterly disgraceful".

Mr. Rooney concludes, in robust fashion:
"In my more bitter moments I see its approach to the wildlife issue as having been a most impressive exercise in creative prevarication. The excuses for not doing anything have been most imaginative".

The reflections by the ISG on their trial protocol from the ISG in public are somewhat inconsistant. While they assert that it was 'their protocol', when pressed on the specifics, the answer is somewhat different.
Professor Bourne to the EFRAcom (18/06/07)
The Irish "...have less welfare considerations than we forced to give to the trapping that we carried out on the trial"

'Forced? John Bourne? Exactly who had him in an armlock, one wonders? He continues;
"Very importantly, there was 100 % farmer co operation and we did not get that.The other thing is that there is no badger group in Ireland".

and:
"Let us go back to 1999 when we started our work. It was made very clear to us by ministers of the day - and they have not refuted it since - that elimination of badgers over large tracts of countryside was not an option for future policy".

A horrified Geoffrey Cox, MP intervened "Is it not the function of science..." but the diminutive Prof. was in full spate. He was not to interrupted and continued:
It was on that basis that we (note that we - ed) designed the trial. We also had to take into account welfare considerations with respect to culling used, and limitations on culling with respect that cubs were not killed or died underground [ ] Those were clear political limitations that we operated under; I have no reason to believe that those political limitations have changed".

Geoffrey Cox, MP then asked Bourne to clarify the report's findings and its conclusions in the light of his statement describing a political steer in what should have been a scientific exercise. Professor Bourne replied thus:
"We repeatedly say "culling, as conducted in the trial." It is important [that] we do say that. Those limitations were not imposed by ourselves. They were imposed by politicians."

and he concluded the session:
"At the end of the day I think you have to accept that it is the price society puts on a badger. [ ] In this country there is a price on a badger and on badger welfare".

An astounded Mr. Cox, whose constituency houses some of the worst hotspots in the country replied:
"I beg your pardon: they put a price on it in terms of the suffering of families, and the slaughtering of tens of thousands of cattle?"

Professor Bourne, now condescendingly patiently explains to Mr. Cox, the difference between 'science' and 'political science':
"Whatever has driven that I do not know [ try copious multi million ££ donations to political parties? - ed) but the fact is that a price has been put on the badger in this country which related to the way we were able to carry out our scientific work. That is exactly what we report".

So there we have it. As the fallout from this expensive farce continues, one may safely assume that Professor Bourne and the ISG, having ignored Krebs original protocol and its clear warnings, grabbed a politically skewed methodology, conveniently labelled "This is the answer, now phrase the question".

That it was an 'exercise' in prevarication is beyond doubt, but to label it 'science' in the same breath as reading how it was steered, is stretching the imagination.

Monday, September 03, 2007

An epitaph to the RBCT

We have been alerted to a recently published comment piece in the broad sheets, which describes the folly of leaving one's brains behind while acccepting exactly what any computer programme extrudes, regardless of input.

As an epitaph to the hallowed computer modelling which fuels so much of the politically 'in the difficult file' decisions which have taken over veterinary and epidemiological policy in recent years, we reproduce it below, with thanks.

It is taken from an article by Tom Stevenson in the Daily Telegraph last week on this summer’s upheavals in the stock markets. He takes issue with the claim that “highly remunerated rocket science can be a short cut to above-average returns with below-average risk”, and continues:

“The two are as incompatible as they always have been and no spurious cocktail of calculus and Greek letters can change that fact…The admission by the biggest and most powerful investment banks that their ‘quantitative’ rocket scientists got their sums wrong undermines the claim…that ‘number crunching is leading decision makers to make different and, by-and-large, better choices’… Garbage in garbage out. If you lend money to people who cannot afford to borrow it, no statistical model or clever repackaging of the debt will make the resultant waste any less toxic. There are no short cuts…[This] kicks away support from the false idea that human experience and judgment are less important than a terabyte-crunching, mechanistic view of the world.”


The ISG final report is full of 'assumptions' and 'hypotheses', but its conclusions, built via a 'simple' mathematical model, on the back of these are now described as 'hard scientific data'. Yet the consequences explored in the report, of various policies are apparently based on the assumption of "equal risk" given to "roughly the same importance" of:

" all sources of infection for cattle, local infection for example across farm boundaries, infection from animals bought in, particularly but not only, from high incidence areas, .... and infection from wildlife, especially badgers."[ 7.24]

In the case of almost every farm, on onsite veterinary officer from local AHO could, in a couple of minutes discount two of those "rough, equal risk 'hypotheses'" and for any badger cull to proceed, should have to do so.

We liked the final paragraph of Tom Stevenson's piece, and particularly his conclusion, describing statisticians modelling of unsustainable debt which, after it had made passage through a mathematical abacas, indicated above-average returns with below-average risk. The fallout from this, its consequential share value switchback ride has finally kicked away "support for the false idea that human experience and judgment are less important than a terabyte-crunching, mechanistic view of the world.”

Not at Defra it hasn't. And if you change the stuff about the consequences of lending to people who can’t afford to borrow, to killing cattle without dealing with all sources of infection, this would make a perfect epitaph to the RBCT Badger Dispersal Trial.