Open access

I was really impressed by the very smooth process through which my paper on cost-effectiveness analysis in the presence of structural zero costs (which I've already mentioned for example here and here; the related R package is described here) has gone in Statistics in Medicine.

The review process was quick and helpful (at least for one of the referees and the editor $-$ the other referee was kind of not very helpful at all). More importantly, thanks to a recent agreement that UCL have signed with various publishers, the paper can be published in open access $-$ the final version is freely available here.

The only thing I'm not quite completely sure I'm happy about is this: like many journals, SiM has a LaTeX template that you can use while preparing your manuscript. The draft version looks quite neat. But then, when they produce the final version, the maths fonts look more like those obtained using an old version of MS Word. Now, I can see that not everybody uses LaTeX, but once they go to the trouble of re-typesetting the accepted papers anyway, I don't see why they don't do it in LaTeX (which would look much better!). 

But I may be missing some trivial point here. In fact, I was talking to somebody from Wiley earlier today, but I forgot to ask...

Christmas came early (or who's the geekiest in the family?)

By pure accident (honest! I didn't do it on purpose!), last week I opened a package that had come in the post for Marta. Too bad that, of all the packages I could have opened, this was the Christmas gift she had bought (in fact I should say manufactured) for me.

There was no point in pretending I didn't know what it was and so we decided I should open it. Now, the question is: who is the geekiest in the family? Is it me for loooving the present, or Marta for even conceiving it?

Death of a statistician from Surbiton

This morning I heard from Christian the news that Dennis Lindley has passed away, last Saturday. He was 90. I think I'd already linked to this video in a different post, but I'm re-linking it right now.

Last (talk before) Christmas

OK, so first things first. I've always: 
1) been a huge fan of the reindeer hat; 
2) wondered what George Michael had actually given as a present to the other guy (if anything!). 

On a much less serious note, as I managed to get (nearly completely) out of teaching for Term 1, I've been able to spend a bit more time on preparing and giving talks this year, particularly since September. On Sunday, I'll present the paper on Bayesian cost-effectiveness analysis for structural zero costs (a preliminary version of the full paper is arxived here, but the paper has also been accepted for publication in Statistics in Medicine and will be open-access shortly $-$ I think I'll talk about this in a separate post) at the 6th International Conference of the ERCIM WG on Computational and Methodological Statistics.

This will be my first time at ERCIM, so I'm not sure what to expect $-$ the session (ES19, room B33) looks quite interesting, although topic of my talk is probably a bit at odds with the others. Anyway, I've given this talk already a few times and for different audiences (eg here) but this time it will be a bit different, because of the changes to the model that I've made in response to the referees' comments and also because of the much shorter time $-$ only 15 minutes this time. 

Update on Bayes Pharma 2014

This is the logo for the 2014 edition of Bayes Pharma. Very soon we'll advertise the programme and more details $-$ I'll post about it as well. 

At the moment, we're finalising the plan but I think we've decided to structure the workshops in this way. As a general rule, we consider three main areas for invited talks. These are:
1) Bayesian methods in clinical trials
2) Advanced Bayesian modelling
3) Bayesian methods for exploratory and epidemiological analysis

Each year, we could find something specific to fit under these (very loose) headers. For 2014, these are the proposed topics:
1) Power prior vs hierarchical modelling
2) Structural Equation Models
3) Health economic evaluation
(this last topic is not surprising $-$ although it was Emmanuel to suggest it, not me!).

Of course, In addition to these we'll have the contributed session (we'll ask people to submit their abstract).

Newest release of BCEA

Very shortly, I'll upload the newest release of BCEA, my R package to post-process the output of a (Bayesian) health economic model and produce systematic summaries (such as graphs and tables) for a full economic evaluation and probabilistic sensitivity analysis (more posts on this are, in random order, here, here, here and here).

I've made changes of two kinds: the first one can go under the header of "cosmetic changes" (like Pietro Rigo, who taught me a couple of courses back in my undergraduate studies, used to say). Basically, I decided to print out the whole script for BCEA and while I was going through it I realised that in several points I wasn't really being very elegant or particularly effective with my code. This was not a huge problem, I think, because speed of execution is generally not an issue $-$ the inputs to the functions are usually moderately small and the operations required are not too complicated. But, once I realised that, it bothered me that my code was full of loops and unnecessary lines and so I changed it. As it happens, the gains in terms of computational speed are not huge (because of the reason I mentioned earlier). But the cockroach lemma still applies...

The second type of changes could be labelled as "substantial" and involve the following. 

1. I've included a function multi.evppi, which implements the method for multivariate analysis of the expected value of partial perfect information, described here. (The next bit of information is totally irrelevant, but when I went to Edinburgh for the MRC clinical trial conference, I caught up with Mark Strong, who gave a talk on this, which motivated me to finish off the script).
2. A utility function called CreateInputs, which can be called to produce the object containing the matrix with the parameters simulated using the MCMC procedure (using JAGS or BUGS) and a vector of parameters (as strings) that can be used to perform the EVPPI analysis.
3. I've also coded up (borrowing from Mark's original scripts) a function to perform diagnostic analysis on the assumptions underlying the Gaussian process model that is used to estimate the EVPPI. That's called diag.evppi.
4. A function that performs structural probabilistic sensitivity analysis, which is called struct.psa. I think this was long due, since basically most of the inputs were already there $-$ you only have to run the model using different specifications and save (some of) the results to a list and then this function will compute the weights to be associated with each model specification (according to the methods specified in this paper by Chris Jackson et al.

This will be version 2.0-2 $-$ I've toyed with the idea of moving up a gear in the numbering, since it seems to me to be a substantial improvement (more for the inclusion of the new functions than anything else). But I also thought that it would be good to have it out there and see what people think of it and possibly test it (I know that some people are using BCEA so hopefully we'll get some feedback in time for our book).

Lost

The results of the ISBA elections have come out and unfortunately, I've been beaten to the post of programme chair for the Section on Biostatistics and Pharmaceutical Statistics.

I am not sure by how much $-$ I meant to ask for more details, but I've been a bit busy last week and didn't have the time. I suspect that the fact that only current members of the sections were allowed to vote wasn't good for me, as, probably, it increased the incumbent's advantage. [CORRECTION: As per ISBA's official numbers, I actually lost 10-9]

Anyway, I should congratulate Telba Irony, who's been re-elected and hopefully get her to take some of my programme points (starting with sponsoring the next BayesPharma) on board anyway.

My talk at the LSHTM

Yesterday I gave a talk on our RDD project at the Centre for Statistical Methodology of the London School of Hygiene and Tropical Medicine. While presenting me, Karla (the organiser of the seminar) joked that I should go for a hat trick of presentations at the LSHTM, since only last month I gave another talk (on the structural zero problems in health economics $-$ on a related note, the paper, which I also discussed here, was actually accepted by Statistics in Medicine).

The main point of this talk was to try and point out various advantages of including genuine prior knowledge in the RDD framework, to try and get suitable estimates and make the assumptions underlying it more robust. I think we need to clarify a couple of points, but also I got good comments, so it was very helpful!

The slides of the talks are here.

PSMR 2014

Registration for the short course on Practical Statistics for Medical Research (PSMR) 2014 are now open. Here is the advert we've published on the BMJ with all the relevant information and details and even more info & details are here $-$ in case you're interested...

BCEs0 version 1.1 on CRAN

As I was responding to the points raised by two referees and the editor on my paper on cost-effectiveness with structural zeros (the preliminary version was here, while I have presenting it in a few talks and discussed it here, here, here and here).

I have to say most of the comments I received made a lot of sense and were extremely helpful. In particular, when I was thinking about how I should address them, I realised that I was much better off by modelling all the prior distributions on the scale of the mean and standard deviation of the cost variable, rather than using the original scale (e.g. rate & shape for the Gamma distribution).

This is true in general, of course, but it is quite helpful in this case, because I want to impose a very informative prior for the subjects for whom a 0 has been observed (so that in the posterior the mean cost is identically 0, a fortiori). I have updated the software webpage, which now reports the full list of inputs required by the main function bces0.

In particular, I have modified the original code so that:

1. a treatment-specific threshold for the default Uniform prior on the mean and standard deviation of the costs for the non-null component (previously, I was assuming a single value to be applied to both treatment being compared);
2. a "robust" option, which by default is set to TRUE, which implies that "minimally informative" Cauchy priors are specified on the coefficients for the pattern model for the zero cost indicator. If robust is set to FALSE, then BCEs0 will use a vague Normal prior instead;
3. a "model.file" option, which allows you to specify the name of the .txt file to which the JAGS model code is saved. This is not quite fundamental, but as I was testing the package I kind of got annoyed that every time I run it, it would overwrite previous versions of the model file, which I may need for future tests. And so I changed this.

I think the paper in its current (hopefully final!) version looks much better and the more I think about the overall problem and how the model deals with it, the more I kind of like it. But then again, as we say in Italian: "ogni scarrafone e' bello a mamma sua", which poorly translates into English as "every cockroach is beatiful to its own mother's eyes"...

Loophole

I think I should thank Marta (again!) for this post, as she made me think about it while we were riding together to the Stan workshop, in one of our now ("A XY", that is, as opposed to "B XY" when we used to do so all the time) rare joint outings on the Vespa.

Lately, quite a few London Buses advertise electronic cigarettes, which we found peculiar, given the ban on tobacco advertising. Now, of course, technically, electronic cigarettes have nothing to do with tobacco, so, I'm guessing, they are perfectly within the law in advertising them.

However (and I must say I don't really know enough about this!), it appears that some evidence is present to hint at potential risks to health due to e-cigarette consumption. So, one may wonder, why are these allowed to advertise without formal investigations on their safety at least planned? Again: I may be completely ignorant of government-commissioned studies into this matter (in which case, well done UK Große Koalition!). But I may also be guessing well, right?...

Imperialstan

Despite the map here, I'm not going to talk about yet another fraction of the former Soviet Empire which is taken the form of a people's republic, possibly with witty British Ambassadors.

In fact, I'm going to talk about the Stan workshop that I have be to, earlier today, which was held at Imperial College. My friend Lea organised it and Mike Betancourt (who's actually in my department at UCL) run the show (brilliantly, it has to be said).

In the morning, Mike gave a brief overview of MCMC and introduced the basics of Hamiltonian Monte Carlo (I think this by Radford Neal is just a great introduction to the topic). Then in the afternoon he concentrated on Stan and rstan in particular (which, unsurprisingly, is the R interface to the actual HMC engine).

I think this was kind of the first of a potential series of similar talks/workshops and I found it very useful. Of course it's always difficult to strike a balance between how in depth you want to go with the theory and the examples, so for instance, I think a little more on the actual NUTS algorithm would have been helpful $-$ but as I said, I know full well how hard it is to do this, so well done, Mike!

Keynote speaker

Earlier today, I was trying to finish preparing the poster for the Clinical Trials Methodology Conference $-$ I'll have both the poster presentation (on the Expected Value of Information under mixed strategies) and my talk on the Stepped Wedge design on the Monday, so by 3pm I'll be just wandering around the sessions having done my duty.

Luckily, XY slept a couple of hours, so I could actually do some work. But then he woke up and while he was playing in the living room, he found my badge from the Chemometrics Workshop and the latest copy of Significance. 

I did put the badge around his neck the first time, but he quickly learned to do it himself; at which point he kept putting it on, while carefully reading the magazine. I think he quite looks the part, doesn't he?

Rescue remedy

Interesting day, today. I woke up really early (3.45am) to catch my flight to Amsterdam to give my talk at the Chemometrics Workshop. The cab got me to the airport early enough so that I could clear security, have a coffee and slowly make my way to the gate. 

But when I arrived there, I realised that before leaving home, on the spur of the moment, I decided to take a different jacket $-$ you know, just to stir things up a little. Too bad that I had left my passport in the jacket that was at home... But, no matter; after all I still have my Italian ID card $-$ I don't often use it, but surely that would come to the rescue and allow me to board my flight. Except that it was kind of expired (well $-$ "kind of" meaning expired two years ago). And so they kindly escorted me outside the departures area.

At which point (it was about 6am), I decided it was time for desperate measures and called home. Thankfully, my sister-in-law Sara is visiting, so, I thought, if I can get Marta, she could come and bring me my passport and XY won't be alone at home. So all will be fine. Except that, of course, at that point I had missed my flight. But they assured me that the next one would still get me to Amsterdam on time. So I just needed to pay a reasonable "recovery fee" and get the passport in time to re-do the check in, re-clear security (and possibly re-take a coffee).

I can't really blame her, but it was kind of difficult to get hold of Marta and I had to ring quite a few times before she actually picked up the phone. And of course, she was quite upset too! (You know the kind of sleepy voice that basically sounds like: "I would probably kill you and then leave you if you were here"). But again, I don't think I can really blame her... Anyway, thankfully, enter super-wife to the rescue! She drove the 45 minutes from our house to the airport and managed to get me my passport on time.

Oh, did I mention that the flight they told me would get me to Amsterdam on time was actually full, so they couldn't rebook me on that one? Yeah $-$ that happened too... Luckily, just when I was about to give up, I thought I'd check with British Airways; I was half expecting they would ask charge £100,000 but in fact I only had to pay just a little more than the "recovery fee" would have been, so I bought a one-way ticket. 

In the end, I did manage to get to Utrecht on time. The talk was scheduled for 11.15 and I got to the conference centre with well over 10 minutes to spare! Because I didn't want to miss on any of the possible excitement of the day, I also witnessed a robbery on the train from Schipol $-$ an American guy got his backpack stolen by a lovely-looking (but clearly evil-acting) lady when the train stopped at one of the stations along the way.

By the way, the talk went well, I thought. It was probably a bit too long and I had to cut one of the examples short, but people seem to have liked it and asked me quite a few questions. I'm at the airport now, waiting to board my flight back home $-$ assuming I still have a home, that is. I did buy chocolate in the hope of placating my lovely, forgiving other half...

Bayesian Biostatistics 2014

This has to do with the ISBA Biostats section (I suppose it will be even more, if I am elected to the post of Program Chair, but I'll try and be involved even if I don't win!): the next Bayesian Biostatistics Conference has just been announced and will be held at MD Anderson (Houston, TX) this coming February (12-14).

I went to the 2009 edition, which was good. Most of the emphasis, as is obvious given the research interest of the organisers, was on clinical trials, in particular early phase. At that time I wasn't really working on that topic, so it was good for the academic experience, but it didn't have immediate impact on my personal work.

Now I am doing some work that is related to this, so this would probably be even more interesting (although it does happen in the middle of the teaching term, so I'm not sure I'll be able to go...). And in any case, the topics are quite diverse:

• Imaging analysis
• NGS data analysis
• Subgroup analysis
• Enrichment designs
• Integrative genomics
• Deconvolving tumor heterogeneity
• Biomarker-driven adaptive designs
• Drawing inferences from large databases
• Anti-infectives and other non-cancer therapeutic areas

so it wouldn't necessarily be just about clinical trials!

Typos in BMHE

No matter how many times you check and no matter how good the publishers are, I'm guessing there's no way out of getting typos in a publication, especially if it's a relatively long one, such as a book.

I've just discovered a couple in BMHE, which I thought I should try and correct (or at least point out) here.

1. Page 45, figure 2.5: the bottom part of the decision tree shows the possible outcomes when the operation does not go well. However, the headings for the branch out of the "Patient lives" random node are both "No". The top one should read "Yes" (much as in the top half of the tree).
2. Because of the changes in the R2jags package, the code in chapter 4 and 5 should be slightly modified to say attach.jags() instead of attach.bugs() $-$ I've already explained this in more details here.
3. Page 186: I don't really like that the transition matrix overflows to the very bottom of the page $-$ but that's not my fault... My original file didn't have that.

I'm sure there are more $-$ I'll post them when I catch them (or, more likely, when people make me notice them: I have to thank my student Shabai for noticing the mistake in the decision tree).

My talk @ the Dutch Chemometrics Symposium

For same reason, Paul Eilers really liked the talk I gave on INLA at the BayesPharma workshop earlier this year and so he invited me to talk at the Dutch Chemometrics Symposium.

Now: you may ask what have I got to do with chemometrics. And you would be right in doing that $-$ in fact, such is the extent of my ignorance on the subject, that I had to actually look "chemometrics" up... But Paul asked me to talk about Bayesian statistics in general, so that was very much up my street, which is why I gladly agreed.

The workshop is this coming Thursday (7th November) in Utrecht. I've put the slides of my talk: "Come to the dark side: we got cookies! An introduction to Bayesian statistics" here.

Fellow me

Last summer I have applied for a NIHR Research Methods fellowship. Earlier this week the results have come out and they have liked my proposal, which is of course great news. 

The idea of this project is to critically evaluate the stepped wedge design (SWD) in clinical trials. This is a relatively new design, effectively an extension of cross-over design, in which a given intervention is rolled out in clusters that unilaterally switch treatment at different time points. The first time point usually coincides with a baseline measurement where all the clusters are assigned to the control arm. Subsequently, clusters begin to receive the active treatment, but, unlike in a standard cross-over trial, once the intervention is given, it is not removed. The time at which the intervention is started is randomly determined.

This on the one hand typically increases the duration of the study (because several time points are usually needed to reach a fixed level of statistical power); however, on the other hand, the SWD has shown the potential to be more efficient than standard cluster randomised (CR) designs.

But of course, much as for standard cross-over designs, the actual gains depend on specific settings and parameters specifications (eg in terms of the the number of clusters and time points; the clusters size; the level of correlation between measurements in the same cluster and across time). So we'll try and investigate these issues and see under which conditions the SWD works better than other strategies. As part of the proposed outputs of this research, we have indicated that we'll produce a toolbox (in R) to perform sample size calculations and guide the analysis of the actual data.

The Big Bayes theorem theory

While we were eating a forkful of what was supposed to be a frittata, but turned out to be very fluffy mushroom scrambled eggs earlier, we were half watching an episode of The Big Bang Theory. 

Long story short, my eye was caught by Sheldon explaining how he is estimating his life expectancy, clearly using Bayes's theorem (although he didn't refer to it in his speech to Leonard).

Good stuff, Shelly!

Bad teacher(s)

This morning there has been some frenzy on the UK media (eg here or here) after the publication of a pamphlet by David Willetts, a junior minister for University and Science under the infamous coalition government.

The minister's point is that, comparatively to what used to be case in the past (notably in 1963 before changes in policy that led to increase in the number of university students), the proportion of time spent teaching by university lecturers is decreased in favour of the time that they spend otherwise.

Now, of course, this is not necessarily bad or good per se, but the minister says in his paper that "Looking back we will wonder how the higher education system was ever allowed to become so lopsided away from teaching." 

Well, one easy answer is of course to point out that apart from the huge increase in the number of students $-$ it would actually be interesting to have reasonable figures on the time spent teaching per-student, in comparison with pre-1963! $-$ the government(s) have switched the emphasis to research by decreasing the amount of funding available for universities and rewarding private initiative to obtain research money, eg from industry, or simply making the process of funding increasingly competitive!

Again, not necessarily a bad thing, but certainly not something to coolly swipe under the rug...

R2jags & BCEA (& the examples from BMHE)

Recently, Yu-Sung Su and Masanao Yajima, the developers of the R2jags package, have released a new version (the current one is 0.03-11). As far as I understand it, one of the main changes is that since the update, R2jags no longer depends on the R2WinBUGS package (although it "imports" it).

The consequence of this is that you can no longer use the R2WinBUGS functions, such as for example bugs or attach.bugs(), by just loading R2jags. In fact, there's a new function attach.jags() that allows you to attach the object you obtain as a result of the call to the jags function and containing, among other things, the MCMC simulations.

More importantly, if you also use BCEA and try to replicate the examples I describe in BMHE (for example see here, here, here and here) you are in trouble. All the code I have produced was running OK under the previous version of R2jags, but now you do get an error message when you try to attach the JAGS object using the attach.bugs() command.

Fortunately, this is not a huge problem and you actually have two options to solve it: the first one is to add to all those scripts a formal call to R2WinBUGS, eg library(R2WinBUGS). This will make the attach.bugs() command available again and so the rest of the code will run OK.

The second way is to actually use the attach.jags() command directly. In this, you don't need to load R2WinBUGS; however (because, as Sheldon Cooper would say: "what's life without whimsy?"), in this case you have to change the argument to this function, since attach.jags() takes a rjags object, while  attach.bugs() wants a bugs object.

So, for example, assume you have the following code.
library(R2jags)
model <- jags(data,inits,parameters.to.save,
        model.file="some_file.txt", n.chains=2, 
        n.iter, n.burnin, n.thin, DIC=TRUE, 
        working.directory=working.dir, progress.bar="text")
and you want to make the object model (and all the elements contained in it) available to your R session, you can either do
library(R2WinBUGS)
attach.bugs(model$BUGSoutput)
(notice that model is an object in the class rjags, while its element BUGSoutput is in the class bugs), or do
attach.jags(model)
directly.

Le Tour

Today I've given the talk on the model for structural zeros and the related R package BCEs0 for the third time in three weeks (this time it was at the London School of Hygiene and Tropical Medicine).

Le Tour is going quite well, I think $-$ in all three occasions, the talk has been well received. What I think is also interesting is that each time I have received a very different set of questions.

At GSK, people in the audience asked questions on the broader methodology for cost-effectiveness analysis (which they weren't probably very familiar with). In Las Palmas, most questions were about the details of the Bayesian model (for example, the use, or misuse, of the DIC as a measure of model fit and to apply structural sensitivity analysis).

Today, most questions were about the substantial aspects of the economic evaluation. For example, Richard made the interesting point that the model for structural zeros could be turned into a model for "structural ones" in the utility measure $-$ the problem being that sometimes when QALYs are used as the measure of effectiveness, a bunch of patients are associated with a value of 1, which indicates maximum utility. 

This effectively generates a two-component mixture (individuals with utilities in $[0;1)$ and individuals with a utility value of exactly 1). The extension of the hurdle model should be able to do the trick in this case too. This may be a good thing to do for my undergraduate student who will do her project on health economics!

Election night(s)

The 2013 ISBA elections are finally underway! From today (and until November 15th) members will be able to cast their vote for several posts, by simply visiting this webpage.

As I mentioned here, this time around I am one of the candidates, specifically for the section on Biostatistics and Pharmaceutical Statistics (my statement is here $-$ just navigate to the relevant section).

Like any self-respecting candidate, I have already cast my vote $-$ it has always amused me that top politicians are always shown on television casting their vote 21 seconds after the ballot is open. Unfortunately, no TV crew was there to record the operations...

Road trip

Today I had a meeting for one of the projects in which I'm involved. This is a big research grant (in which I have a marginal role, to be honest), with the aim of evaluating some occupational therapy for people with dementia and their carers. 

Nothing special there, you might say; in fact, nothing special there. Except that the meeting was to be held in a far and remote location...

View Larger Map

The main research team are located in a hospital in North East London (well, I suppose that's technically Essex) $-$ a good 32 miles (that's 52 km) from home. So I had a kind-of-nice road trip to get to my meeting, which basically crossed London from one side to the other. Fortunately it wasn't too cold and it hasn't rained, while I was going, so the journey wasn't too bad. Some parts of London are just awesome! 

Self-syndication

This is a piece I've written for The SWITCH project website. SWITCH is a research project addressing issues related to the social market economy in Europe. The topics addressed in the project range from macro financial sustainability conditions to the dynamics of science and innovation. I'm self-syndicating the piece here.

The OECD has just released an interesting working paper comparing the issue of value of pharmaceutical innovations and its impact on pricing. The main, well known argument is that policy on pricing should have impact in the short term (by lowering costs with respect to benefits) as well as in the long term (by encouraging R&D and innovation). The report analyses 12 countries, mostly (but not exclusively) in qualitative terms. One obvious result is that heterogeneity among them is observed, particularly among countries which have guidelines on the use of economic evaluation to drive the process of reimbursement and pricing, and those which do not. 

Of course, even among those formally considering cost-effectiveness considerations there are differences and the evidence is not conclusive as to, for example, which cost-effectiveness threshold should be selected. In particular, this has implications when discussing specific diseases and thus pharmaceutical interventions, such as terminal illnesses (eg cancer). In that respect, there is perhaps an argument to suggest the use of alternative utility functions, which for example include a measure of lower risk aversion on the part of the decision-maker. 

While this aspect is only marginally hinted in the report, the tremendous implications from the technical point of view is clear. Tools such as the expected value of information can be used more extensively to aid in quantifying the value of deferring decisions (on reimbursement or pricing) in order to reduce the uncertainty characterising the evidence presented by the company. This is relevant as, almost invariably, files are based on limited evidence from clinical trials, which may need complementing from post-marketing or observational data. 

Finally (a point missing from the report — although to be fair, this was not their main objective), the all important issue of the underlying assumption that the market is able to adjust instantly to the introduction of new, cost-effective interventions: most of the times, we observe older interventions staying on the market for a longer time. While this can be justified (eg by bringing to bear the uncertainty in the cost-effectiveness profile of the new intervention), it is also likely to produce a loss of optimality. Again, the expected value of information could be used to determine a form of pay back (eg from the companies allowed to remain on the market with a drug which is potentially non cost-effective), which could lead to alternative pricing policies [Baio, G., Russo, P. (2009). A Decision-Theoretic Framework for the Application of Cost-Effectiveness Analysis in Regulatory Processes. Pharmacoeconomics 27(8), 645-655 doi:10.2165/11310250-000000000-00000].

The (third) runway bride

I think I should disclaim the conflict of interest in this one (since Marta is one of the authors of the paper), but it was really, really cool to see her study on the impact on health of noise pollution close to airports in the newspapers today (for example here or here)! 

I thought that the Daily Mail would be also all over the news, while, interestingly, there's nothing on their homepage (although they do mention the article here).

I think the choice of pictures to accompany the articles is also quite interesting: The Guardian chose a rather romantic picture of an airplane taking off from Heathrow at dawn (or sunset $-$ I couldn't quite tell), while both the BBC and the Daily Mail had pictures of airplanes extremely close to properties or the ground (well, they were landing, after all...).

The original paper is linked here.

Happy birthday

Sylvia Richardson (who's now the head of the MRC Biostatistics Unit in Cambridge, and part of our RDD project) asks me to advertise the MRC Biostatistic Unit's Centenary Conference, which will be held in Queens' College Cambridge on March 26th 2014. 

More info are at this webpage, and here is a flyer $-$ the deadline for submission of contributed papers and posters is 28th October.

Video killed the radio stars

Francisco tells me that they have uploaded my talk (which I gave last week in ULPGC). I haven't seen it all, but the bit I did see is not too bad, I thought... Check it out!

 

Nice & weird people in the Canary island (oh: I went there for work too!)

The past one has been a very interesting week, which I've spent visiting the University of Las Palmas, in the Canary Island. Since it was the last week on maternity leave for Marta, we all went. I knew the weather would be good, but we didn't expect it to be so nice $-$ like proper summer weather!

We were really lucky to start with, as while we were driving to the airport last Saturday, I accidentally got the wrong turn before entering the motorway and by doing that, we were able to see a massive queue ahead. So we decided to ask our phone to drive us to the airport through the Surrey back roads. We got a bit scared because the navigator kept telling us to turn into small roads that looked nothing like going to the airport, but in the end we actually got there on time (unlike 15 people who actually missed the flight).

Of all the family, XY has enjoyed the week the most. He has spent most of the time eating, walking around and being chatted up by the nice, friendly people in the streets of Las Palmas. At first he was a bit puzzled by people speaking in Spanish (I guess it sounded close enough to Italian, but not quite Italian, so it freaked him out a bit). But then he got used to it and started waiving at people when they were stopping us in the streets to say that he was so "precioso", or "lindo", or "guapo".

Speaking of which, I found it amazing how we could relatively easily get by with our Spanitalian. In fact, I did study a bit Spanish, which was helpful, but for most things we could speak to people without resorting to hand-waiving or English (of course that's not ground-breaking news $-$ but it was nice).

Which brings me to the actual reason of the trip, which was a week of meeting with Miguel Negrín and Francisco Vázquez Polo. We mainly talked about some work in health economics, specifically related to the application of utility functions different from the standard net benefit (which is defined as $ke - c$, where $e,c$ are the variables of clinical effectiveness and cost, respectively, and $k$ is a willingness-to-pay parameter, used to rescale the benefits and put them on a monetary scale). In the beginning I wasn't quite convinced about their argument, but then we got to talk and, while I'm still not 100% sure of all the implications, I think this is an interesting problem and definitely worth investigating. So we'll try and work on this together, which is a good outcome of the visit $-$ and I'll post about this, once we (I) have clearer ideas/material. I also gave a seminar on the structural zero model, expanding on the talk I gave at GSK a few days back. I think this is really interesting and I seem to get useful comments, which I'll definitely use in my revision of the paper.

The last day we rented a car and decided to tour the island of Gran Canaria. We'd never been there before, so we didn't really know much, but Miguel and Francisco made a couple of suggestions. While we were walking around in Maspalomas, a weird Italian lady stopped us and started to blab. She looked interested in us, but in fact asked us how old we were 3 times in the space of 5 minutes and told us pretty much the same things about 20 times. As it turned out, she was working for a holiday resort (apparently they have taken some sand from Bahamas and brought it there before building the resort $-$ I am still annoyed with myself that I didn't turned around and left as she actually said that, as if it was some amazing thing which was supposed to make me want to divorce Marta and marry her instead). Anyway, she wanted us to get some sort of scratch'n'win ticket. We weren't really interested but she started scratching a couple and apparently we won either an iPad, a free one-week holiday (in one of their "amazing resorts"), or €500 in cash. To get the prize we "only" had to go and see the resort with the Bahamas beach, which we could only do by going with a cab that she had to call (and there was no other way, "because those are the rules"). We went back to the car and actually visited the nice little town of Teror (not very big and not very much to do, but still quite nice and much less hot!).

You stole my idea!

Earlier today, Gareth has showed me a recent, interesting paper by Michael Sweeting (and colleagues). In the paper, Micheal et al describe their work on a R package to extend on the framework of the Continual Reassessment Method (the original paper by John O'Quigley and colleagues is here), which is a particular design that can be used in dose-finding studies.

No one has really stolen anything (just to be clear!) but I did have the same idea just a couple of months ago $-$ Gareth was working on a similar problem and asked me to have a chat about how one can apply the CRM, which is essentially a Bayesian method.

In its standard form, essentially, the CRM uses a very simple model based on very simple priors (typically exponentials or Gamma). While we were discussing it, I thought it would be nice to expand this and may be make a package that would link to JAGS or BUGS and allow you to select the prior from a wider range. Which, as it turns out, is just what Michael et al have done! 

Well $-$ at least I can see that was a good idea!

My talk @ GSK

This Thursday I'll give a talk at the GSK Statistics Forum. Erika (with whom I shared a train journey to the 2012 BayesPharma and a group walk in Oxfordshire a few years back) now works at GSK and invited me. I will talk about the model for cost-effectiveness with structural zero costs, on which I am working at the moment.

In fact, I managed to get hold of some new and more interesting data, which I will use to produce a better, more articulated example (although I think for the talk, I'll stick with the original acupuncture example, which is part of the working paper). 

I'll post my slides later, in case anyone's interested...

Parum PI

A couple of weeks ago, the MRC-funded research project on the Regression Discontinuity Design (of which I'm the Principal Investigator) has officially started, so I thought I wrote a few lines of update about it, after the couple of posts (here and here) referring to presentations we've given on (very preliminary) work we've done.

Unfortunately, I don't think I can quite claim to have the physique du role to be a Magnum PI [hence, and to show off that I did study Latin in school $-$ although you may argue that I could have easily used Google Translate... but I haven't: promised! $-$ the title of the post], if only for the fact that we're having a terrible September, weather-wise, here in London...

Let me be clear that I'm not complaining and of course I am very happy that we got the grant. But I must say that being PI is at the same time a very exciting and exhausting role (I'll pretend that it hasn't occurred to me that the project is just started). Today I put on my most Hawaiian shirt and spent most of the time trying to sort out a few admin things and sending emails. 

Hopefully, we'll shortly have something a bit more substantial to report about $-$ the signs are all there, luckily...

Attendance in parliament (in Italy)

Earlier today, I found some interesting data on Italian MPs voting records in the current parliament (which was opened last April). The data are available for both Houses (the Italian system has a lower house, called Camera $-$ effectively the same as the House of Commons in the UK $-$ and an upper house, the Senate). The Camera is made by 630 MPs, while there are "only" 321 senators (some of whom are appointed for life by the President of the Republic). For each MP, data are recorded on their political affiliation, constituency, number of votes attended and total number of votes in the current parliament. I thought I tried some relatively simple model to see whether there are some substantial differences in terms of party or constituency. 

I think the two houses can be modelled separately, at least as a first approximation. So, for example, say I concentrate on the lower House and define $y_i \sim \mbox{Binomial}(\pi_i,n_i)$ where for $i=1,\ldots,630$, $y_i$ is the number of recorded attended votes, $n_i$ is the total number of votes and $\pi_i$ is the individual probability of attendance. [NB: not all MPs are "exposed" to the same number of votes, as some have been appointed after by-elections, ie after parliament was opened last April. Also, the system does not record the reason for absence and so some of them may be legitimate (eg illness, or due to other institutional engagements). Nevertheless, the data give some good indication of the actual activity of the MPs].

I used a very simple model to define the attendance probability 
$$ \mbox{logit}(\pi_i) = \alpha + \beta_{p_i} + \gamma_{c_i}$$where the index $p_i$ indicates the party to which the $i-$th MP is affiliated and the index $c_i$ indicate their constituency. Of course, parties and constituencies will be replicated within the dataset (because MPs cluster within them). Also, other individual- or party- or constituency-specific variables may be relevant to explain away the different attitudes to participating to the parliamentary work $-$ but I'll keep the model simple, mainly because I'm a bit lazy and am not spending time to find these other variables!

The model is completed specifying vague prior for the overall average attendance rate (on the logit scale), $\alpha$ and structured priors for $\beta_j \sim \mbox{Normal}(0,\tau_p)$ and $\gamma_k \sim \mbox{Normal}(0,\tau_c)$, with $\sigma_p = \frac{1}{\sqrt{\tau_p}} \sim \mbox{Uniform}(0,10)$ and $\sigma_c = \frac{1}{\sqrt{\tau_c}} \sim \mbox{Uniform}(0,10)$. 

With this specification, the coefficients $\beta_j$ define the incremental effect of the $j=1,\ldots,N_p$ parties and $\gamma_k$ is the incremental effect of the $k=1,\ldots,N_c$ constituencies on the propensity of each MP to attend the votes. Negative values for the coefficients means that a given party/constituency decrease the chance of attendance.

I ran this model and here're the results, in the form of coefficient plots, reporting the posterior interval estimation for the effects $\beta$ and $\gamma$. For the Camera, the MPs affiliated with the "Movimento 5 Stelle" party and those with "Sinistra Ecologia Libertà" show positive propensity to participate in votes, while FdI ("Fratelli d'Italia") and PdL ("Popolo delle Libertà", Berlusconi's party) are substantially associated with negative propensity to attend.

Constituency also seem to have some differential effect; MPs from Valle d'Aosta, Liguria, Sardinia, Friuli Venezia Giulia, some provinces in Lombardia (but not including that of Milan) and some provinces in Sicily (including Palermo) have positive propensity to attend the votes.

In the senate, the situation is kind-of-different, if only for the fact that by the nature of the Italian system, some parties are not even the same. All the "positive" party effects disappear and no party is significantly associated with higher propensity of attendance. However, "PdL" and "Scelta Civica" (the party of former Prime Minister, Mario Monti) do show substantially negative values, indicating their senators have a lower propensity to show up at parliamentary votes.