I believe it is claimed of that great mathematician GH Hardy that when he came up with his famous little result in the probability of genetics (what we now know as the Hardy-Weinberg law) he was so worried about the sneers he would suffer from his fellow mathematicians at Cambridge if they found out, that he chose to publish it in an American general science (as opposed to mathematics) journal. Times change, however, and although I avoid paying any attention whatsoever to JAM, because as I get older , I find that my appetite for fiction diminishes, I just know that some of my fellow statisticians are bound to read it.
Now I have always made it a golden rule to avoid collaborating with any medics in publication. The way it usually works, when I work on a project, is that I decide what goes in the analysis plan and am ruthless as to what I won't permit. There is to be no testing of baseline balance as a means of choosing whether to use or ignore covariates, no analysis of change scores, no dichotomising of continuous variables and in particular no responder analyses and no numbers needed to treat, no treating of ordered categories as if they were nominal, no dynamite detonator graphs, and in fact no traces of mean values over time with standard error bars. Above all there is to be no garbage about using log-rank tests rather than proportional hazards tests 'because they are more robust' or eschewing analysis of covariance 'because the parallel assumption may not be valid'. And when it comes to cross-over trials, there is to be no testing or indeed adjusting for carry-over.
Such is the prejudice of medical journals in favour of all the garbage I won't permit, that this means that it is more or less impossible to publish a paper based on my analyses. This usually means that additional analyses have to be provided either by Pannostrum or by the 'tart' we have got to be the lead author. You may ask why I am allowed to get away with this. Well that's a long story and I don't feel like going into it just now. Anyway, the net result is that we have published but vastly inferior versions of the trial report for many of the trials I have worked on without my name on them and a series of methodological didactic pieces for which I am the sole author illustrating the follies of common approaches to design and analysis. And that's the way I like it.
But now it turns out that my name will be on a paper that not only appeared in JAM but has a responder analysis in it. The embarrassment could not be greater than if I was a minister of the cloth who was discovered to have membership of a lap-dancing club. So how did I get my name on this publication' Oh the shame, the shame. It seems that due to some new directive regarding authorship, our medical publication department have made it mandatory that every paper in addition to the usual sprinkling of external tarts must have at least one medic and one statistician from Pannostrum on the author list. A recent high profile trial of ours involving cardiovascular safety of anti-asthmatics has now completed. Angina Cutter and Harvey Puffer are the Pannostrum medics on the list. I thought that I could safely assign one of the data-slaves who worked on the project to be the stats co-author but it seems they want a more senior person on the list. I should be flattered. I am not. Anyway I was fully determined to tell Harvey that the answer was 'no' and that I was not about to break a long habit of collaboration in which he published the analyses I told him not to and I published a paper explaining what garbage his analysis was.
Now, in fact, Harvey and I had had a run-in quite recently on another matter and this is the subject of another of the three humiliations to which I referred. You may recall that Harvey had put it down on his Personal Development Profile as one of his personal goals to learn more about statistics. Well he has gone one better and he has started teaching it also or rather arranging to have it taught.
It was actually Angina who alerted me to it. She rang me up the other day. 'Hallo Guernsey,' she said, 'I am getting a bit confused about a statistical concept and I just know that you will be able to sort me out.'
'She's a smart girl', I thought and not for the first time.
'Well actually it's P-values. I think you had always taught me that a P-value was the probability of the data given the hypothesis and not the hypothesis given the data.'
'Well that's nearly right,' I said, 'strictly speaking it is the probability of a result as extreme or more extreme than that observed given the null hypothesis, but you are on the right lines. What you are describing is the likelihood.'
'Right,' she said, 'or rather it isn't. This MUSE course I have been doing has it described as the 'probability that the null-hypothesis is true.'
'MUSE?', I said. (Well I didn't actually say the capital letters but already I was expecting an acronym.)
'Yes. It stands for Making Understanding Statistics Easy. Surely you have heard of it. All the medical advisors are doing it.
Human resources commissioned this company Psystistics to produce the course. It has ten one hour sessions. Harvey is the Pannostrum liaison on the project.'
Of course I checked it out and of course I knew it would be garbage before I even looked at it. It turned out to be worse than I thought. Pages and pages of very smart looking nonsense under the strap-line, let our MUSE inspire you to better statistics. Example 'Question When should you use a within-patient study? Answer Whenever you want to see how subjects change over time. For example, in a treatment for obesity you will obviously want to see how much weight each subject loses, so a within-patient study is best.'
I confronted Harvey at the earliest opportunity. 'Harvey, What possessed you to become involved with this piece of garbage?', I said, diplomatically.
'Now Guernsey. Don't give me that not invented here attitude. Psystistics are very good at what they do. They are very qualified in their field.'
'I have looked them up,' I said. 'They are a small company offering, according to their rather flash website, to help you Use the power of the internet to unleash your personnel's potential. Amongst their staff I recognise two psychologists, three computer scientists and a small mixed bag of media studies and marketing people. That doesn't constitute expertise in statistics in my book. Look at the appalling mistake they made about P-values'
'Angina told me about that. You really are being very pedantic. I don't see what is wrong about defining it as the probability that the null hypothesis is true.'
'It's the error of the transposed conditional,' I said. 'The probability that a randomly chosen woman is suffering from breast cancer is low. You will think I am quibbling if I insist that it does not follow, therefore, that the probability that a randomly chosen breast cancer victim is a woman is also low.'
Harvey looked puzzled. 'I don't know what you are talking about,' he said. 'Obviously those two probabilities are very different. But the P-value must be the probability that the null hypothesis is true. That's what I learned at medical school.'
Anyway the net result of this is that war is declared by me on our completely useless human resources department for commissioning this nonsense in the first place. Watch this space.
The third humiliation occurred on a flight over to see our transatlantic colleagues in Medicine Springs. I was travelling business and sitting, as is my wont, at the bar. A rather attractive young lady was polishing off what was obviously the latest in a long series of vodka tonics and we got chatting. What did she do? It turned out she was an actress going across for some auditions. What did I do? 'I am a mathematician', I lied.
'Oh. Does that mean you have a degree in mathematics?'
'Yes' I said, lying again.
'Well in that case, can you explain something I have always wondered about? How is it possible to have a degree in mathematics. Is there enough to study?'
I suddenly realised the depth of ignorance out there as to what a subject like statistics involves. Of course we were talking about mathematics but any ignorance the public may have about what mathematics involves applies a fortiori to statistics. This young lady had presumable learned to add subtract and multiply at school. Admittedly that would still leave the mystery of division but surely a university degree could not be just about that. So what do these three stories have in common? A complete contempt for the quantitative amongst the numerically challenged is the answer. If I could persuade my co-authors to present statistics as they should be presented, if I could make my colleagues accord statistics the respect they would grant immunology, if I could make the general public have some understanding as to what it is that mathematicians, let alone statisticians, do, how much sweeter my life would be.
I was discussing the JAM business with my wife and she opined that it was all a lot of fuss about nothing. Surely the whole point of medical research was to have it published and nothing was ever perfect.
I replied that that might be so but I still could not understand how I had allowed Harvey to persuade me to be a co-author when I was usually so obdurate.
'Was it Harvey?,' she said. 'Are you sure it wasn't Angina?'
'Oh,' I said, 'perhaps it was Angina.'
'I thought so,' she said smiling.
'There's a very important factor involved.'
'What factor?' I said.
'The little finger factor.'