Of course, a Bayes 500 would be even better and more powerful (although more fuel-consuming...). But: the best we can afford (strictly for time constraints $-$ time since publication, that is) is a Bayes 250.
This is what promises to be a very, very good conference to celebrate the anniversary of the publication of the "Essay towards solving a Problem in the Doctrine of Chances" by the Reverend Thomas Bayes, in the Philosophical Transactions of the Royal Society of London.
The latest number of Significance actually published a paper arguing for a legitimate recognition of Richard Price's contribution to the paper (I think there's possibly more than something true about this, although I thought that the article was trying too hard to make the point).
Anyway, the programme of the two-day conference (to be held on 19-20 June at the Royal Statistical Society in London) is quite impressive! I'm planning to go, although the fee is a bit high $-$ perhaps I'll try to see if I can get some money from random savers across Europe...
Yesterday I was "interviewed" by Panagiotis Petrou, who is working within the EU project Adapting European health systems to diversity (ADAPT). [He did suit up for the "interview", while, because it was in my office, I wore my normal work clothes. On second thought, I probably should have done a Barney...]
We had been in contact via email for a health economic model he was working on (and for which I informally suggested a few things) and then he asked me if I could share my thoughts on the issue they are trying to tackle.
As I understand it, their main objective is to think of how to deal, from the health economic point of view, with ethnically and culturally diverse sub-populations who may not have rights to access to public health systems around Europe, but who end up being treated anyway in accordance with the principles of universal access.
That's an interesting problem and of course it has quite clear health economics impact, because these sub-groups of the population might end up accessing high-cost emergency treatments for conditions that may have been controlled using routine service (eg GPs). But in many cases this is not possible, because they are "out of the system".
But the problem is more complex than that, because before you can even think of the health economics of it, there's the whole political issue to be accounted for $-$ different countries even within the EU may have different approaches to how "aliens" are treated and who pays for that.
Perhaps it has quite a few links with works being developed to study unauthorised migrations (eg this and this). I did mention this aspect to Panagiotis and given that both streams have connections with the EU it may even be easier to combine them. Or not?
We finally got around to prepare everything we needed to advertise the position which will be available in the MRC grant we've been awarded last year.
The project will run for 30 months and we're looking for a post-doctoral candidate to work on the Research Discontinuity Design as applied to the evaluation of the effect of interventions (mainly pharmacological) in primary care.
I think the project is quite interesting because it is both about developing some theoretical advancements on the causal inference literature and making some analysis on real-practice observational data. Similarly, it cuts across several disciplines and has ramifications bordering on health economic evaluation, epidemiology and statistics. That's why the research team is so big.
All the relevant info is here. The deadline for applications is 14 April 2013 (which, incidentally is the day after my cousin's birthday $-$ but of course that's not why we chose it!). We plan to shortlist and interview by the end of April and then start as soon as possible. And before you ask, no: candidates will not have to build their own chair...
As every year, come April our group hold a short course on Practical Statistics in Medical Research. I think this has run for several years now and it's reasonably established. The course is aimed at health care professionals (ie non statisticians) and we teach various aspects of experimental and observational studies, methods for statistical analysis of medical data and try to give a general feel for what a non-statistician should know (about statistics, that is...).
So far, I have not been too involved (as I only teach one class, plus a couple of tutorials/practicals), but, apparently, I'll organise it from next year, courtesy of some sort of rotation policy in place among the people in the group. I'm counting on the fact that the machine is well oiled and, to a certain extent, most things are kind of in place.
I think registration is not closed yet $-$ have a look here if you're interested!
Today I spent the morning reading a PhD thesis that I need to examine (scheduled for next month) and the afternoon marking in-course assessment (ICA) papers for my course (Social Statistics). Neither activity is the most amusing in the world (although thesis reading can be interesting, some times more than others, of course)! Also, the weather is quite crappy, so these were as good as any other excuse for not leaving the house this morning.
The PhD thesis is about metabolomics. That's really not my area, but because the candidate has used Bayesian methods to analyse their data, some of which similar to what I normally work with, the supervisor asked me to examine it. I haven't got to the part where the model is actually discussed yet, but I have to say I am still not too clear about what metabolomics really is and whether it can be really the game-changer that people sometimes make it to be (of course, this is not necessarily the candidate's fault...).
As for the ICAs, I'm really torn. I have asked the students to find a real social survey, read about it (searching for as much information as they possibly could) and then report about the main characteristics (target population/study population/sampling frame and design, etc).
On the one hand, I think it's good for the students to do this kind of exercise. Many get it right and give good answers, but many are quite confused about basic stuff (it is not uncommon that somebody selects something like a census and report it as if it were a survey). So, because the ICA usually is just a marginal part of the overall mark that the students get, it is helpful to face these problems before the exam. However, on the other hand, marking the papers becomes a very, very tedious exercise for me. If I had to decide today, I would definitely scrap this kind of question for any future exam/ICA I will write.
But perhaps I should sleep on it $-$ which I will if I don't take a break now!
I've received an email by one of the contributors of statisticsblog.com and, while we were both hanging out of our parents' back, we've decided we wanted to become friends.
I had encountered this blog a couple of times in the past and their approach is quite similar to mine (so I instinctively like it; that's not to say that I do like everything that is similar to me, but it certainly makes it easier to get a good feeling out of it...).
Anyway, in particular, I kind of like the fact that they have a motto (In Monte Carlo we trust) $-$ it made me think that perhaps I should get a motto of my own.
That has happened to me once before, while Marta and I were spending some time in Boston during our PhD. At the time, she was working on statistical analysis of genomic data and so she was based in a lab, which was led by an Italian (nay, Florentine) researcher. This one time, I went to see a presentation and on the first slides, he had modified Harvard's motto "Veritas" to a much more convivial "In vino veritas". But then again, he did come from Chianti...
Last year Julien invited me to the Bayes-Pharma workshop in Aachen, which turned out to be quite nice.
This year, I've been invited in the scientific board for the new edition, which will be held in Rotterdam this coming May. So far, we've had a very good response, with many people sending in their abstracts (but I think there is still time to do so!). We have, I think, a very interesting programme lined up.
I'll give a short introduction to INLA but, perhaps even more than that, I'm quite looking forward the session on Bayesian methods for missing data on the second day. Nicky Best and Alexina Mason (both from Imperial College) will present their work, which is based on a fully Bayesian approach.
This contrasts with more "standard" methods, like multiple imputation (MI, eg as in this), which are Bayesian in nature, but which people normally apply and evaluate in terms of their frequentist properties. While I think that the general framework upon which MI is built is quite neat, I have always found it a bit confusing that it starts as a Bayesian procedure, only to them move towards something not Bayesian at all.
I don't really know the history of this all too well, but I guess this may be an indirect effect of the fact that when MI methods were originally developed Bayesian computations were still very difficult to make.
I'll post more on this as the actual meeting approaches.
