If I'm invited to give a talk at a scientific or statistical conference, and I don't have any new results or stuff to talk about (or that I'm able to talk about), I have a standby talk that I dust down and deliver, which is an unpublished monograph, really, and the outline of a book I'm supposed to be writing with two friends, only work keeps getting in the way. The talk is called 'Evidence and Belief', and I use it to set out a paradigm for research in those areas of science where the stuff we don't know still outweighs, massively, the stuff that we do.
My thesis is quite simple. Evidence for, and inference about, parameters which describe some sort of population 'thing' of interest (like the clinical effect of an addictive substance on a population of its users) is properly summarised through something we (statisticians) call a likelihood function. Then your prior belief (prior, that is, to seeing the evidence) about that parameter of interest is properly modulated, in the light of the data (the evidence), via application of the Bayes' theorem to that likelihood function. But neither evidence about a parameter, nor belief about it (or related hypotheses) are the same thing at all as a decision to act on the basis of that evidence. Bayesians (the name for people who think like I do) think that risk (of a bad outcome) exists only at the point of the decision to act; not at the time when inference about the parameter of interest is carried out.
You can tell where this is going.
You might assume, given this faith I have in a demarcation between evidence, belief and action, that I'd support the government's peremptory sacking of Professor David Nutt. He was supposed to produce evidence, and to make inference about the relative dangers inherent in drinking to excess vs popping an E (showing my age there, and, before you ask, no, never). But he was certainly not responsible for making a decision to act. That decision should be made by the relevant minister alone.
But David Nutt wasn't sacked for forcing a private member's bill into parliament and attempting to change the law. He didn't even campaign for the replacement of Alan Johnson as Home Secretary with our own spokesman, Chris Grayling, who thinks, er, exactly the same thing as Alan Johnson, except that he adds that Alan Johnson should have sacked Professor Nutt even more quickly (before he reported his scientific investigations at all, presumably).
Professor Nutt was sacked because he kept repeating the inference which the evidence he collected had forced him to carry out: that the criminalisation ranking of addictive substances is not in a one-to-one, strictly increasing mapping with the true level of risk to an individual who indulges in those substances. Indeed the two rankings quite lack concordance. Addictive substances which are freely available, such as alcohol, have a much higher risk to their users' health than does, for example, ecstasy.
That's a fact. Sacking David Nutt won't change it. Alan Johnson (and depressingly, our own spokesman) can make any decision he wishes, of course, but he can't pretend to have discovered a new law of pharmacology, or a new epidemiological finding, which subverts the work of Professor Nutt.
As it happens, I don't actually have a strong view of the classification of illegal substances. I think the pharmacology and epidemiology are both compelling, but I don't care to have alcohol criminalised because lots of people can't handle it, and I don't think the man-made ranks of categories of legal sanction have much to do with the quantity of skunk which infests our estates in Hackney. I'm not in favour of liberalisation. I don't share Professor Nutt's passion for change in the slightest.
I think it was wrong to sack David Nutt, not because of his views, but because the Home Secretary has abused his position in doing so, and is undermining the contribution of researchers to government. If the only acceptable opinion from an independent (of parliament) scientist is the one that the Home Secretary already holds, what's the point of asking for it in the first place?
We've been here before, of course, though the tabloid press have forgotten their outrage the last time the government attempted to manipulate statistical evidence in order to force the headline it wanted to see. It was about knife crime, that time, and the qualms of the NHS statistician about the spin Jacquie Smith put on the data were quashed. Correctly, George Osborne wants to make the Statistical Service totally independent of the government; presumably a decision with which Chris Grayling does not agree.
