Sunday, 27 April 2014


My friend Natalia has posted a link to this blog on her Facebook profile. Basically, this is the story of a paper published in 1994 in the medical journal Diabetes Care

As it turns out, this paper "discovered" a method to: 
1) determine total area under a curve with precision; 
2) calculate area with varied shapes that may or may not intercept on one or both X/Y axes; 
3) estimate total area under a curve plotted against varied time intervals (abscissas), whereas other formulas only allow the same time interval; and 
4) compare total areas of metabolic curves produced by different studies.

Now, if you have so much as taken some basic calculus at some point in your life, you should realise that all this is actually done using the "trapezoidal rule", one of the basic ideas in calculus and integration.

The paper is relatively old (at least it dates back to an era prior to the massive development of the internet), but I think it is still quite bad that it got through peer review without nobody at Diabetes Care bothering to ask somebody outside their pool of expert reviewers (and with some mathematical skills) to check that the method was actually sound and innovative.

I've checked on the Web of Science (at the MRC Centenary workshop last month, Robin Henderson joked that people use it to check the number of citation of somebody else's papers $-$ but they use Google Scholars to check their own): the paper has 187 citations. What's even worse is that many of these are actually quite recent (as far as earlier this year)!

(I think the story is not new and quite a few people have already commented on it, for example here).

Friday, 25 April 2014

Bayes Pharma 2014 - final programme

Here's the finalised programme for Bayes Pharma. (Of course I may be biased), but I think it is quite interesting and we have a wide variety of topics, so there should be plenty to get your teeth into! We still have some places available (and as always, students can register for free!).

Day 1 – 11 June 2014
9.00 - 9.30. Registration

Invited session 1. Bayesian Method in Clinical Trials: Priors from Historical Data
9.30 - 11.00. Beat Neuenschwander (Novartis)
11.00 - 11.30. Coffee break
11.30 - 12.30. David Dejardin (Bristol-Meyers-Squibb) & Emmanuel Lesaffre (Erasmus University Rotterdam)

12.30 - 13.30. Lunch break

Invited session 2. Advanced Bayesian modelling: Structural Equation Models
13.30 - 14.30. Peter Congdon (Queen Mary University London)

14.30 - 15.00. Coffee break

15.00 - 16.00. Phil Woodward (Pfizer). Advantages of a wholly Bayesian approach to assessing efficacy in early drug development: a case study

Contributed session 1
16.00 - 16.30. Haifang Ni (Utrecht University). Use of power priors to incorporate imperfect information in the analysis of veterinary clinical trials
16.30 - 17.00. Astrid Jullion (Novartis). Using mixture priors for robust inference: application in Bayesian dose escalation trials 

Social Event
17.30 - 19.00. Trip to East London on a double-decker bus to visit Bayes tomb in Bunhill Fields 

Day 2 – 12 June 2014
Invited session 3. Bayesian methods for exploratory and epidemiological analysis: Health economic evaluation
9.00 - 10.00. Chris Jackson (MRC Cambridge)
10.00 - 11.00. Richard Nixon (Novartis, Basel)
11.00 - 11.30. Coffee break
11.30 - 12.30. Gianluca Baio (UCL)

12.30 - 13.30. Lunch break

Contributed session 2
13.30 - 14.00. Nicky Welton (University of Bristol). Extrapolation of trial-based survival curves: constraints based on external information 
14.00 - 14.30. Mohsen Mohammadzadeh (Tehran University). Bayesian Analysis of Spatio-Temporal Dynamic Panel Models with Fixed and Random Effects 
14.30 - 15.00. Katrin Haussler (University College London). Bayesian health economic modelling of different human papillomavirus (HPV) vaccination strategies in a static and pseudo-dynamic setting: the BEST II study

15.00 - 15.30. Coffee break

Contributed session 3
15.30 - 16.00. Yingbo Wang (Imperial College London). Using Propensity Score to adjust for unmeasured confounders in small area studies
16.00 - 16.30. Eric Rozet (Arlenda). Transfer of analytical methods: the Bayesian way 
16.30 - 17.00. Bayoue Li (Erasmus University Rotterdam). Joint modelling of a multilevel factor analytic model and a multilevel covariance regression model

Day 3 – 13 June 2014
Contributed session 4
9.30 - 10.00. Paul Meyvisch (Janssen). Optimizing sample size in relative bioavailability trials using a Bayesian decision theoretic framework
10.00 - 10.30. Foteini Strimenopoulou & Ros Walley (UCB). Clinical and pre-clinical applications of Bayesian methods at UCB
10.30 - 11.00. Moza Al-Balushi (Sultan Qaboos University). Determinants of Contraceptive Use in Oman: A Hierarchical Modeling Approach

11.00 - 11.30. Coffee break

Contributed session 5
11.30 - 12.00. Jean-Fran├žois Michiels (Arlenda). A Bayesian framework for conducting effective bridging between references under uncertainty
12.00 - 12.30. Leandro Garcia-Barrado (Hasselt University). Incorporating post-development study information to reduce the sample size of diagnostic biomarker validation designs
12.30 - 13.00. Kazem Nasserinejad (Erasmus University Rotterdam). Bayesian growth mixture models to distinguish hemoglobin value trajectories in blood donors

13.00 - 13.10. Farewell 
13.10 - 14.00. Lunch

Books & roses (& chatty taxi-drivers)

I knew that Wednesday was el dia de San Jordi but I hadn't realised the way in which Catalans celebrate it.

My flight from Heathrow was slightly delayed, so by the time I arrived in Barcelona it was past 9pm. Still, on my way from the airport I had registered the $-$ I thought $-$ unusual number of women carrying a rose with them. But I was pretty tired, so I didn't really think much of it.

In the morning, I caught an early train to Girona and then a taxi to get to the Universitat. The driver was very, very chatty and she wouldn't stop speaking, so I had to switch to Spanish-mode to the full of my (limited) abilities $-$ in fact, I was quite happy and impressed with myself for being able to sustain a long conversation. 

Anyway (enough self-patting on the back!), she explained the tradition and told me that for San Jordi men are supposed to give their loved ones a rose, while women should give theirs a book, which I thought was quite nice (especially the book thing $-$ I suspect that as far as the roses are concerned, it must have become something like Valentines cards. But with a book it's probably become a much less tacky & commercial activity?).

Friday, 18 April 2014

My talks @ Universitat de Girona

Just after Easter, I'll go for a very quick trip to lovely Girona, where Marc Saez has invited me to give two talks.

The first one will be a re-run of the short course on INLA that I did at Bayes Pharma last year. It's scheduled (and prepared) as a 3-hours talk, in which I briefly introduce the general Bayesian problem, quickly glance over the basics of computation and then spend some time on INLA, describing the theory behind it and then presenting a few examples on how you use it for inference. I've modified the talk slightly and changed a couple of examples.

That will take up the morning and after an half hour break I'll give another seminar; this time I'll talk about the structural zero model. I have presented this a couple of times before, but I think in all cases it was before I finalised the model. So this time the talk should be better.

Wednesday, 16 April 2014

The Granville incident

Earlier this morning, there was some commotion on the allstat mailing list (if you don't know what it is, that's a UK-based discussion list specifically focussed on statistics; it's been active for quite some time and usually you get useful information on mostly, but not exclusively academic vacancies, conferences, etc).

Vincent Granville is a self-described "visionary data scientist" who often sends emails to the list $-$ most of which the list does not seem to receive with pleasure... [NB I believe that for today, Granville has sought and received more than his fair share of publicity, so I won't link to his webpage $-$ you can google him if you're interested]

Anyway, in his email today, Granville has openly offered a 250 dollars reward to write a review of his book on the Amazon page. In fact, he said he would reward the four "best" reviews (it is not clear what "best" means in this case $-$ I suspect it's to do with how enthusiastic they are...).

This has of course unleashed a reaction of anger and indignation among the users of the list (although Michael Bretscher from Imperial College noticed that "ironically, if all corrupt activities were communicated so transparently, the world might be a better place...").

I agree with the general sentiment and I think that Granville is coming from a very, very different place than basically everybody else who's active on the list. Some people has suggested banning him altogether, which I can sympathise with. But even more simply, is this not a classic situation in which your spam filter becomes your best friend?...

Monday, 14 April 2014

Bayes Pharma 2014 - nearly there...

We're nearly done with (most of) the preparation for Bayes Pharma 2014. We've received quite a few abstracts for the contributed talks $-$ many different topics but in general very interesting work, I thought.

We've managed to secure several sponsorships $-$ we kind of look like one of those weird football shirts where they stick up as many brands as they possibly can. But it was nice to involve ISBA and the RSS (in addition to the Quetelet society, who have a long standing association with the conference).

I've nearly finalised the social event (I'll post some gossip as we get closer) and all I'm left to do is to arrange lunches/coffees et al. There are still some available places $-$ all the information needed for registration is here. As always, students go free!

Causal Inference in Health, Economic and Social Sciences

The programme of the forthcoming UK Causal Inference Meeting "Causal Inference in Health, Economic and Social Sciences" is just out. The short conference will be at the end of the month (28th and 29th of April) at the University of Cambridge.

I indirectly feature as Aidan (who's part of our RDD team) is giving a presentation in one of the sessions. His talk is entitled "The Effect of Prior Beliefs on Causal Effect Estimators within a Bayesian Regression Discontinuity Design" and basically comes out as a follow up to our paper.

The idea is to implement the RDD within a full Bayesian framework and to actually assess carefully what's the gain of doing it this way. As is often the case, in some situations, there may be unwanted impact of the prior on the causal estimators, although generally there are advantages (both computationally and in terms of including extra sources of information $-$ that is crucial, especially when you want to go beyond statistical analysis, in my opinion).

Sunday, 6 April 2014

Seven (a-day)

This week (among other things, including my Vespa breaking down twice in three days) I was busy taking part in an interview panel for a research associate position, together with colleagues in the Medical School at UCL.

One of the questions we were asking to the candidates was about commenting a new study (incidentally, by researchers at UCL) which using data from the Health Survey for England argued that the current "optimal" regime of consuming 5 portions of vegetables and fruit per day could (should) be in fact increased to at least 7, to reduce risk of death.

The point of the question was of course to get the candidates to recognise the possibility of confounding $-$ of course people consuming more veg & fruit might have a much lower risk of death to start with, due to different life-style, etc. (A few candidates got it right straight away, others less so).

But I think this has even more interestingly (eg, economic) implications in terms of the actual applicability of the health policy, in its current form as well as in terms of potential modifications, like this article in today's Guardian (which I thought was spot-on!) suggests.