Many years ago at dinner I was sitting next to [shameless name-drop alert] Elliot Richardson. We were talking about the limits of the proper roles for government, and he suggested that one of the few legitimate activities government could engage in was the mass collection and provision of fundamental data, using the powers to collect data that only government ever really has. UK Biobank, though now essentially privatised under Oxford University, might be though of as one such public endeavour.
For those who've never heard of it, it was set up 20 years ago, with private sector as well as public sector stakeholders from the start. But its links with the NHS are so central to its work, I think it's fair to describe it as a public venture. 500,000 initial volunteers in the age-range 40-69 went through a day's initial screening on all manner of bodily parameters, with occasional follow-up studies subsequently (via questionnaire) and - here's the critical bit - ongoing tracking of their subsequent medical histories from the NHS. It so happens I was one of the randomly-invited pool[1], and have followed developments with interest. Obviously we are dying off a bit over time (and emigrating beyond the reach of the NHS), but the extant pool is still very large.
Bona fide scholars from anywhere in the world can apply for access to the gargantuan data-set, with details of their proposed study, large numbers of which have been conducted. A while back I attended a presentation on what can be achieved with such a large database, and it's impressive. In terms of statistical significant and validity, the difference between a sample of (say) 10,000 on a particular issue is greater than two orders of magnitude better than a sample of 100.
The latest project has been to conduct very thorough whole-body and cerebral scanning of 100,000 participants, coupled with lengthy cognitive testing on the same day, and two weeks of heart monitoring thereafter. This is scanning on an industrial scale, and the efficiency of their process was impressive: I've just taken part in that myself. It was very interesting, and some of it fun, albeit that sadly you don't get given your cognitive results.
The whole thing is highly regarded world-wide, as well it might be. A few years ago the Chinese mooted doing one of their own, but so far as I know there's been no follow-through. This is the kind of thing we (still) do uniquely well.
ND
_________________________
[1] I say 'random', but of course the pool was designed to be a representative cross-section of the UK population.
1 comment:
Those of us wot have done Stats for a living, even on a minor scale, might choose to say "pseudo-random". It probably doesn't matter much since the more gormless part of the nation now seems to use "random" to mean "haphazard".
Mind you when you are learning stats it's frustrating that "random" in "random variable" has a different meaning from "random" in "random sample". God knows why they hadn't cleaned up their terminology decades ago.
Post a Comment