Saturday 25 April 2015

SPDEVPPI

We've just arxived our paper on efficient computation for the Expected Value of Partial Perfect Information (EVPPI) based on SPDE-INLA. The EVPPI is a decision-theoretic measure of the impact of uncertainty in some of the parameters in a model on the final decision, informed by current evidence. Basically, it measures how much one should be willing to pay to gather new information to reduce uncertainty in the future outcomes. As such, it can be used to prioritise research and thus has been advocated as a very useful tool in health economic evaluation. Trouble is that it can be very difficult (or even impossible) to compute analytically and even a simulation approach may require an impracticable number of simulations.

The paper is part of Anna's PhD and in it we draw heavily on previous work on Gaussian Process (GP) regression to compute the EVPPI done by Mark Strong and colleagues at Sheffield. Our main idea is to express the GP regression model for the estimation of the EVPPI using a "fictional" spatial problem. As we show in the paper, this allows us to make use of clever models which simplify the computations and we can use INLA to obtain the results in a super-fast way.

As we show in the paper, our method can compute the EVPPI in a matter of seconds and this is virtually irrespective of the number of parameters involved. More importantly, there's a very large reduction in the computational time, in comparison to other non-parametric methods. Also, the accuracy of the estimation is very good.

We're implementing this into the next release of BCEA, which we plan on having ready in a matter of a few weeks!

Monday 20 April 2015

There's no I in Mutual

Today I've received an email from a prospective PhD candidate, who says that he (or she) would like to do his (or her) PhD under my supervision on a topic of mutual interest. Except his (or her) interests are in Physics or Applied Mathematics.

I know I'm being a bit mean here and that there's no reason why physicists or mathematicians can't work on Bayesian statistics. But this person actually goes on to state that their expertise is in Rheology and Dynamics of Complex Fluids e.g. particle suspension, droplet and polymer suspension and various other multiphase flows. That I find it a bit more complex to qualify as a set of mutual interests (to mine!)...

Friday 17 April 2015

Stata goes Bayesian

The other day, my colleague Gareth pointed out a very interesting piece of news. The new version of Stata is just out. Now, I'm not a super-Stata user (although I think it's a good package), but the interesting news is that they have now developed a specific module for Bayesian analysis.  

This was probably coming $-$ I've recently reviewed a new book exploring the integration of Stata and BUGS, pretty much similar to the R2OpenBUGS package for R (there are in fact a few similar packages to interface MCMC software such as JAGS or Stan or WinBUGS to R). The new version of Stata can now skip this step (by and large) as many "more or less" Bayesian models have been hard-coded and can be fitted using standard Stata commands. 

I'll check this out!

Monday 13 April 2015

Conflict of interest

Disclaimer: I'm fully aware of the obvious conflict of interest here, but also I think that this looks really good, so I'll write about it anyway.

This post is to highlight that Marta's and Michela's book on Spatial and Spatio-temporal Bayesian Models with R - INLA is finally out (I think it can be pre-ordered although it will be officially available early in May). I think the book is really good as it describes the underlying theory of INLA and makes the effort of presenting a unified framework, including examples and R code

The only downside to this, now that the book is sort-of out of Marta's mind, is that she'll go back to architect-mode and start again to suggest new ways in which we can move the furniture (or even worse, the rooms) around. I should find another topic for her to write a book, soon...

Thursday 9 April 2015

Health economic combat

A couple of weeks ago we decided to create a more formal website for our research group within the department of Statistical Science at UCL. 

The group includes the PhD students involved in health economic-related topics (basically all under my supervision, although for all of them the help of my colleagues who act as second supervisors is being invaluable!) and some of my colleagues in the department. 

The website collects all the relevant information, including a description of the current research projects (mainly at PhD level, but also for the MSc projects and, potentially, for Undergraduate projects), as well as details of our seminar series. We'll also link to some (forthcoming) publications and news.

I've fought (and just about won) to resist the temptation of using the acronym Sheeva, as in Statistics and Health Economic Evaluation, for the group $-$ I guess it would have really been too geeky...

Tuesday 7 April 2015

Unsolicitors

This is probably just me being a bit grumpy, but I guess this happens to many people. I have just received an email (and it's not the first time) from a random scientific journal (this time it's a medical journal) inviting me to publish my research.

Except that my research has nothing to do whatsoever with the topics that this journal is interested in. Also, I love how invariably these emails say something like: "If possible, I would appreciate receiving your submission by" tomorrow (or a date very very close in the future) and usually also say something like: "Please respond to this mail by
" today.

At least this one wished me "a nice and healthy day ahead!!"...