Number of the Month

March   2005

TRUTH

Truth will come to light; murder cannot be hid long.
Merchant of Venice

Every now and then comes along a book that throws a searchlight beam on the nonsense and iniquities of the age. Such was Le Fanu's account of the state of decline of scientific medicine in 2000. Now a highly authoritative book has appeared that puts a bomb under some of the most cherished tenets of the environmental movement. It is by a retired professor of organic chemistry from Wrocław. In telling the true story of a family of organic compounds, it exposes the chicanery, mendacity and sheer callous inhumanity of the quasi-religious orthodoxy that has seized control of the media and the political stage across the world.

It is destined to be ignored by officialdom, but all adherents of science and its methods should feel duty bound to read it and shout about it.

6/03/05

Temporary note: Number Watch is experiencing difficulties in sending e-mails at present.

And apologies for absence: Your bending author has been suffering from those Ol' Bill Gates' customer blues, which have now been cured by Outlookectomy.

From the scaremongers' manual

Correspondence from some of the more ancient of number watchers, following last month’s comments about NSAIDs, coupled with newspaper reports that many women are still avoiding the benefits of HRT, led to your bending author feeling a new law coming on.

It started out as a law of new therapies, but after consideration it seemed to be of wider application, just as the law of league tables applied even more effectively to targets. Receivers of both NSAIDs and HRT have reported to Number Watch how these therapies gave them their lives back, yet they both come under strident attack by scaremongers. A nasty old cynic might ascribe the motives of these people to their desire for fame and grants, but where do they get the bullets to load their blunderbusses?

It is all based on three of the fallacies that have been fully covered in the books associated with this web site:

  1. The independence fallacy:
    It is quite possible for a wholly beneficial therapy to be proved to increase deaths from various causes ( a numerical illustration is given in Sorry, wrong number!, p201). Because we all die of something, reduction of deaths by one cause inevitably leads to an increase in deaths from all other causes. Thus the  beneficial therapy gets all the adverse headlines.
  2. The data dredge (extreme value) fallacy
    Many of the scaremongering studies do not involve the rightly condemned observational type, but are properly conducted double blind trials (with an exception noted below). Where they stray from rectitude is in looking at a number of potential diseases, yet applying the statistics as though each one were being conducted separately. Thus, to simplify grossly, a positive result is claimed if some threshold is crossed that has been assigned a 95% probability of being significant and not caused by normal random variation. That is to say that it is calculated a priori that probability of the boundary not being accidentally transgressed is 0.95. Fair enough, but if you are looking at two independent diseases, the probability of neither transgressing is (0.95)2, or 0.90  and in general for N diseases it is (0.95)N. When N is 5, the probability is 0.77, and for N=10 it is 0.6.  The 95% boundary is thought by many outside epidemiology to be inadequate, but 60%? We are not often told how many diseases were actually in the data dredge, but each is usually treated as though it were the result of a separate trial. So the vaunted 95% confidence level is, in reality, a lot less.
    This is yet another powerful argument for a more stringent requirement than 95% (or P<0.05). For, the closer the percentage is to 100, the less effect raising to a power has (unity to any power is still unity).
  3. Premature termination
    The exception to the proper conduct of double blind studies came to light with the Tamoxifen scandal. An international study had been going on for six years, when the Americans suddenly terminated their contribution and announced the results. The excuse was that the interim result suggested that the drug reduces breast cancer by 45%. Apart from the fact that few outside epidemiology accept 0.55 as a significant relative risk, the European scientists were horrified that the experiment had been unilaterally unblinded in this way.
    You might think that the scandal would put an end to such practices, but NO; they have now become the norm, particularly in the case of HRT. As a trial progresses, the relative risks and confidence intervals wander up and down as random walks. One might stray across a boundary, but who is to say that it might not also stray back before the assigned period of the trial comes to an end.  Termination of the study prevents this unfortunate occurrence and secures the all important headlines.
    The other worry is that, if someone is monitoring the progress of the results, is the trial truly blind? There are all sorts of subtle ways that a breach in security could influence the results. For example, what is actually put on death certificates is the choice of the medic concerned and could have varying connections with reality.
    A dramatic and heart rending example of the latter phenomenon occurred this month, when a two day old baby became the latest victim of the MRSA scandal (which gave us last month's number of the month). The family were shocked to find that MRSA was not mentioned on the death certificate and refused to sign until it was changed. The Health Secretary said “Around 2,000 children at a very early age die and on average one of those deaths is caused by MRSA and that is a tragedy.” If doctors are not putting it on the death certificate, how is it going to appear in the statistics?

Anyway, as we were saying, all that led to:

Brignell’s law of beneficial developments:

The intensity of the scaremongering attack on any new development is proportional to the level of benefit that it endows.

Corollary:

Alternative therapies do not come under attack.

22/03/05

Polls apart

Oh, to be rich! To be able to hide away on a remote island until it is all over. Yes, it is general election time in the UK; the season of mock sincerity, dirty tricks, frangible promises, mounting hysteria and, worst of all, opinion polls. Tories close their gap as Blair loses public's trust says the right-leaning Daily Telegraph, observing a one point gap. Poll sees Blair stretching lead yells the left-leaning Guardian, claiming an eight point lead for Labour.

How much variation should we expect just due to random sampling error? To make it perspicuous, let us take a simplified model. Just consider one party and assume that it actually has the loyalty of one third of the electorate. We then have a binomial distribution with p = 1/3. The expected error in percentage points for a given sample size of n can then be stated in the terms of the standard deviation:

which is about 47 divided by the square root of n.

A typical value of n in published opinion polls in the UK media is 1000, so the expected random variation would be plus or minus 1.5%, a range of 3%.

Size of the sample is vital to reliability rightly states the Telegraph, but just what increase in accuracy do they achieve? Unfortunately, they face the law of diminishing returns that is embodied in the square root relationship. Although they question nearly twice the number of electors they achieve only a root two reduction in error, to a range of about two percent. Furthermore, as the Telegraph article points out, there is the question of dependence: as the numbers are constrained to add up to 100, the gain for one party must be matched by a fall for another, so the apparent change is effectively doubled. This effect is analogous to the fallacy of ratios.

We are now in a position to conduct our own poll without the tedium of doing all those interviews. We select three random numbers from the binomial distribution (n=1000, p=1/3) convert to percentages, then (somewhat crudely) model the dependence by assigning to the other party half the remaining percentage:

This was one of three realisations, chosen for artistic merit. It does not take much imagination to create the dramatic headlines – Pinks open up a six point lead over the Blues, then lose it! Yet it is all based on sampling errors alone. Even more sinister is the small print accompanying each survey, reporting “adjustments” made according to unnamed conventions known only to those privy to the mysteries. As we have seen with epidemiology, such adjustments tend to involve questionable assumptions, such as wealth being indicated by postal code. Of course, there are many other factors operating. The two cases quoted above were either side of an event with a whiff of Watergate about it, in which an injudicious speech at a private meeting was secretly taped and passed to opponents, but there is nevertheless a disturbing tendency for pollsters to produce results pleasing to their customers.

Shock horror

The results of the British Farm Trials of GM crops are out and, guess what, they show that these do what they are designed to – allow farmers to grow food crops instead of weeds and pests. Yet the establishment media have seized upon them to condemn this valuable development as the worst thing since the Black Death. The end for GM crops: Final British trial confirms threat to wildlife screams the Independent. VINDICATED - Finally, the long and popular struggle to stop the growing of GM crops in Britain appears to be over yells the Mail.

It would be tedious to go through all the fallacies, non sequiturs and downright lies that are used to bolster these lurid opinions (see the AgBioView Newsletter for all the sordid details) but just take one example. It is observed that there is a reduction of butterflies over GM crops, but they do not distinguish between the pretty ones, that most of us try to attract to our gardens with suitable food plants, and the white ones that are among the most precocious destroyers of food crops. In the strange world of the organic religious movement, it is all right to squash the caterpillars by hand, but not to employ any other method. Chacun à son goût.

28/03/05

Size matters

Bureaucrats like Big. For them only big is beautiful. Whether it is computer systems (and we all know what happened to them) or university research groups, only the overblown will be tolerated. Thus we have a rash of collectivisations in various sections of British society, the latest being in the Health business. The most brutal example is the new contract for pharmacists, which will drive the smaller ones out of business. This is yet another disaster for small rural communities, who have found their pubs and post offices disappearing through direct Government intervention, which diverts the business to suitable large entities like supermarkets and banks. All such facilities play vital multiple roles in the small community, much too untidy a concept for the bureaucratic mind. Anyway, why do people insist on living in inefficient isolated groups? They are not needed in the countryside, now that farmers are being paid to be park keepers. Why can’t people be content to be herded into towns, where all the politicians and bureaucrats live (when they are not weekending in their country homes, which are conveniently insulated from the local communities)?

Likewise General Practioners are to be gathered into more efficient large units. They will be the more “entrepreneurial” medics in “super-surgeries”. No more will old Dr Jones know that Mrs Smith’s problem, regardless of the symptoms she presents, arise from the troubles with her youngest. It will be “next please, take these pills, next please” – so much more efficient.

One of the great myths of our age is the Management Myth, which regards management as an art and a science in its own right, quite independent of the entity that is being managed. It is exemplified by the burgeoning MBA courses. The health service used to be run by doctors and matrons, but they proved inefficient at responding to targets and directives from the Government managers, so they had to be swamped with managers of their own.

Here are some headlines from The Times, not exactly an opposition mouthpiece, over the last five days:

Help me, I'm a doctor, and I can't carry on;  Doctors on warpath to save small surgeries; A disease called despair on the NHS front line; Super-surgery plan signals end for the family doctor.

The greatest irony is that all these political prescriptions are handed down by a minister, but they do not apply to his own constituents. Because of the mangling of the British Constitution by The Great Leader, England’s Scottish New Labour Government not only allows health matters north of the border to handled by their own parliament,  it denies the English the same right, and imposes a Scots MP on them.

Number of the month - 80

This is percentage cut in the number of medical practices proposed by the UK Department of Health. People in small communities will have to get on their theoretical bus route if they require treatment.

All a bit parochial for non-English readers, but take it as a warning.

31/03/05

Index

 

 

moncler outlet usa Moncler outlet hermes outlet prada outlet gucci outlet dior outlet lv outlet chloe outlet