Richard Everitt

With Shreya Sinha Roy, Christian Robert and Rito Dutta, our paper "Prequential posteriors" has been accepted in the special issue on Recent Advances in Dynamical Monte Carlo Methods of Japanese Journal of Statistics and Data Science. link.springer.com/ar...

We use a predictive-sequential loss function to fit deep generative forecasting models for data assimilation, with waste-free SMC for fitting the neural network + Bernstein-von Mises type theorem on the asymptotic behaviour of the generalised posterior, and applications to Lorenz96 and WeatherBench.

lancaster.ac.ukShreya Sinha Roy - Lancaster UniversityProfile page for Shreya Sinha Roy at Lancaster University

Congratulations to Shreya, who now has her first published paper, and has recently started in an exciting new role in the Mathematics for AI in Real-world Systems group at Lancaster.

Ivis Kerama, Tom Thorne and I recently had our paper "Rare event ABC-SMC^2" published in Statistics and Computing. In this paper we introduce an alternative to ABC-SMC that scales better to higher dimensional summary statistics. link.springer.com/ar...

View full thread

In these examples we show how different MCMC moves can be used as part of the likelihood estimator, including slice sampling (as in original rare event ABC paper) and the preconditioned Crank-Nicolson Metropolis-Adjusted Langevin algorithm (in the new Lotka-Volterra example).

The journal also gave us this link rdcu.be/iYUPiZxA95Ka for sharing a view-only version of the paper (not sure if anyone needs this since the paper is open access).

View full thread

The reviewers helped us a lot in developing this paper (thanks to them). Compared to earlier versions, we: (a) provide a heuristic argument that the computational cost of the algorithm scales cubicly with the dimension of the data (compared to exponentially for ABC-SMC);

(b) have a wider range of examples, looking at moving average models to show the method's use in model comparison, a Lotka-Volterra model for illustrating how the SMC sequence of distributions is helpful in exploring the parameter space, and an example of duplication-divergence models for networks.

link.springer.comA rare event approach to high-dimensional approximate Bayesian computationStatistics and Computing - Approximate Bayesian computation (ABC) methods permit approximate inference for intractable likelihoods when it is possible to simulate from the model. However, they...

A key idea is to use a reparametrisation trick, rewriting the simulator model as a deterministic function of the parameter and other random variables. Through this, we make use of the same likelihood estimator as used in the "particle MCMC version" of ABC-MCMC

28 Aug

One year Tutor position in Warwick Stats. Closing deadline: 6th September. warwick-careers.tal....

7 Aug

Two year Teaching Fellow post at Warwick Stats. CLOSING DEADLINE: Monday 17 August 2026 at 11:55pm (UK Time). warwick-careers.tal....

30 June

I'm not sure one is supposed to repost praise, but I hope it can be forgiven in this case, since we had a lot indecision as to what this paper should look like before putting it out. In this situation it feels nice to have a vote of confidence!

Aki Vehtari@avehtari.bsky.social

Excellent paper and I fully agree with their conclusions. We had similar unpublished experimental results but ended up prioritizing other projects. This is a more extensive study of the issue and happy to see it published! @bayesianstats.bsky.social

30 June

New paper with Jia Le Tan (who is at ISBA!) and Nicola Walker. We tried out Pareto smoothing weight-adjustment in SMC, but found there were only minimal benefits since the variance reduction through using a sequence of targets dominates the impact of the adjustment. http://arxiv.org/abs...

11 June

Warwick Statistics is recruiting a Tutor. This is a one year fixed-term role for the academic year 26-27. Closing Deadline: Thursday 9th July 2026 at 11:55pm (UK Time) warwick-careers.tal....

1 June

Alongside our paper "Bayesian Inference of Pathogen Phylogeography using the Structured Coalescent Model", Ian Roberts released the R package StructCoalescent. You can install it using the command devtools::install_github("IanPRoberts/StructCoalescent") github.com/IanPRober...

18 May

To go along with our paper "Bayesian inference of reproduction number from epidemiological and genetic data using particle MCMC", Alicia Gill released an accompanying R package, EpiSky. You can install it using devtools::install_github("alicia-gill/EpiSky") github.com/alicia-gi...

4 May

To go with our Bayesian Analysis paper "Speeding up Inference of Homologous Recombination in Bacteria", we made a version of the ClonalOrigin package that is callable from R, and which contains the improved RJMCMC method for inference. You can find it at github.com/richardge...

github.comGitHub - fjmedinaaguayo/ClonOr_cpp: Clonal Origin model using multiple AIS steps · GitHubClonal Origin model using multiple AIS steps. Contribute to fjmedinaaguayo/ClonOr_cpp development by creating an account on GitHub.

4 May

This is based on Felipe's code at github.com/fjmedinaa... which includes the new methods. Felipe's code is a modification of the original RJMCMC inference for ClonalOrigin at github.com/xavierdid..., was written by Xavier and his coauthors for the paper academic.oup.com/gen...

21 Apr

We are currently advertising for two Teaching Fellows and four Tutors in the Department of Statistics. Deadline on Thursday 23rd April. www.jobs.ac.uk/job/D... www.jobs.ac.uk/job/D...

20 Apr

Felipe Medina Aguayo, Xavier Didelot and I had our work on "Speeding up Inference of Homologous Recombination in Bacteria" published in Bayesian Analysis. We use new parallel annealing-based reversible jump MCMC to obtain more accurate inference for the ClonalOrigin model. projecteuclid.org/jo...

arxiv.orgOn the utility of Metropolis-Hastings with asymmetric acceptance ratioThe Metropolis-Hastings algorithm allows one to sample asymptotically from any probability distribution $π$. There has been recently much work devoted to the development of variants of the MH...

20 Apr

The interesting methodological paper "On the utility of Metropolis-Hastings with asymmetric acceptance ratio" on which this work is based can be found at

27 Mar

We still have a few places at our CRiSM and LMS workshop at Warwick on 16-17th April on "Environmental Modelling meets Infectious Diseases", exploring the links between these two fields. The schedule is now finalised and available on the website http://go.warwick.ac... Visit the website to register.

23 Mar

Leonardo Ripoli and I arxived the paper "Sequential Monte Carlo with active subspaces". Within an SMC sampler we adaptively identify directions in which the likelihood is not informative, and use this to improve MCMC efficiency at each iteration of the SMC. arxiv.org/abs/2411.0...

Richard Everitt reposted

Rafe Meager (they/them)

@economeager.bsky.social

12 Mar

what function is the biggest jumpscare to you. i'll go first, clear choice for me at least it's obviously the gamma function.

9 Mar

Last year Leonardo Ripoli and I archived the paper "Improved MCMC with active subspaces". We identify directions in the parameter space in which the likelihood is not informative, and explore ways of making MCMC more efficient by focussing on the directions where it is. arxiv.org/abs/2501.0...

uqsay.orgUQSay #96The ninety-sixth UQSay seminar on UQ, DACE and related topics will take place online on Thursday afternoon, March 12, 2026. ...

10 Mar

I am speaking about this work at the #UQSay seminar on Thursday www.uqsay.org/2026/02/uqsa...

Richard Everitt reposted

9 Mar

OWABI⁷, 25 March 2026: Robust Simulation Based Inference (11am UK time) Speaker: Larry Wasserman (Carnegie Mellon University) Title: Robust Simulation Based Inference warwick.ac.uk/fac/sc...

artic-network.github.ioPearTree — Phylogenetic Tree Viewer

5 Mar

Pretty remarkable that Bethell's first three hundreds as a professional cricketer were in an ODI, a Test and a T20I (I think).

Richard Everitt reposted

Andrew Rambaut (🐏🎀)

@arambaut.bsky.social

28 Feb

Replying to @arambaut.bsky.social

So I pleased to announce the conceptual spawn of FigTree: PearTree (acronym still to be finalised). If you want to dive right in it is hosted as a web app here: artic-network.github.io/peartree (click the “Example...” button for immediate candy and then click every button you can find).

Richard Everitt reposted

arxiv.orgPrequential posteriorsData assimilation is a fundamental task in updating forecasting models upon observing new data, with applications ranging from weather prediction to online reinforcement learning. Deep generative...

23 Feb

The next OWABI seminar will be given by Shreya Sinha Roy (Warwick), who will talk about "Prequential posteriors" on Thursday the 26th February at 11am UK time. arxiv.org/abs/2511.1... www.warwick.ac.uk/owabi MS Teams link: teams.microsoft.com/...

23 Feb

In July Shu Huang (with me, Massi Tamborrino and Adam Johansen) posted a paper "Inference for Diffusion Processes via Controlled Sequential Monte Carlo and Splitting Schemes". arxiv.org/abs/2507.1...

23 Feb

We use SMC to estimate the likelihood of SDE parameters, combining controlled SMC (to obtain low variance estimators) with numerical splitting schemes and diffusion bridges (to reduce the time-discretisation bias). We show how to use the approach in three different observation regimes.

9 Feb

Ian Roberts published his first paper "Bayesian Inference of Pathogen Phylogeography using the Structured Coalescent Model" with me, Jere Koskela and Xavier Didelot. We describe an efficient reversible jump MCMC method for inference for the structured coalescent. journals.plos.org/pl...

30 Jan

This year the European Meeting of Statisticians will be in Lugano, Switzerland, 24 - 28 August 2026. The deadline for abstract submission is 31 March 2026. https://www.ems26.org/

Richard Everitt reposted

28 Jan

Tomorrow's OWABI seminar by Louis Sharrock (UCL), who will talk about "Sequential Neural Score Estimation: Likelihood-free inference with conditional score base diffusion models" (29th Jan, 11am UK time). Join: teams.microsoft.com/... Meeting ID: 388 990 396 214 49 Passcode: f3Dh22XL

26 Jan

"Annealing Strategies for Variance Reduction in Balance Heuristic Estimators" by Felipe Medina Aguayo and I is accepted in Stats and Computing. Multiple importance sampling uses multiple proposals to reduce variance. We use annealing to significantly further reduce variance. https://rdcu.be/eqe8j

link.springer.comAnnealing Strategies for Variance Reduction in Balance Heuristic EstimatorsStatistics and Computing - The computation of normalisation constants, or marginal likelihoods, is essential in Bayesian inference since they provide information about model adequacy and facilitate...

26 Jan

Publisher link: link.springer.com/ar... Early version on arxiv: arxiv.org/abs/1908.0... Xi'an's Og: xianblog.wordpress.c...

15 Jan

Today is the deadline for contributed talks and posters for our event on Environmental Modelling meets Infectious Diseases. Early-career researchers successful in their application to give a talk will be funded up to £200 for their travel costs. go.warwick.ac.uk/env...

12 Jan

Shreya Sinha Roy's first paper (joint with me, Christian Robert and Rito Dutta) is on "Generalized Bayesian deep reinforcement learning". We use a neural network as a model in Bayesian reinforcement learning, trained using a generalized prequential scoring rule with SMC arxiv.org/abs/2412.1...

*Environmental Modelling meets Infectious Diseases: from Mathematics to Applications* Upcoming workshop in Warwick Stats, supported CRiSM, LMS and RSS. We aim to promote the transfer of ideas between the environmental modelling and infectious disease modelling communities. http://go.warwick.ac...

View full thread

We also welcome submissions to the contributed talk and poster sessions up to the 15th of January. The submission form is at http://go.warwick.ac... Early-career researchers successful in their application to give a talk will be funded up to £200 for their travel costs.

Speakers: Prof. Marta Blangiardo, Imperial Dr Hannah Christensen, Oxford Prof. Sophie Dabo-Niang, Lille Dr Jill Johnson, Sheffield Dr TJ McKinley, Exeter Dr Ben Swallow, St Andrews Dr William Thurston, Met Office

Featured topics will include numerical modelling over a range of temporal and spatial scales, and statistical inference, calibration and uncertainty quantification of model parameters. Dates are 16-17 April 2026. Register for free at http://go.warwick.ac...

New paper accepted in JRSSC with Alicia Gill, Jere Koskela and Xavier Didelot on "Bayesian Inference of Reproduction Number from Epidemiological and Genetic Data Using Particle MCMC". We supplement case data with a dated phylogeny inferred from sequence data arxiv.org/pdf/2311.0...

academic.oup.comBayesian inference of reproduction number from epidemiological and genetic data using particle MCMCAbstract. Inference of the reproduction number through time is of vital importance during an epidemic outbreak. Typically, epidemiologists tackle this usin

JRSSC link now available: academic.oup.com/jrsssc/advan...

We infer prevalence and reproduction number of a disease over time, along with the reporting probability. The paper considers applications to HIV-1 in North Carolina and tuberculosis in Buenos Aires, and compares with alternative approaches that use only epidemiological or only genetic data.

(Realised that I am very behind with posting papers to social media. More to follow in the New Year. Not sure yet how far back in time I will go...)

8 Dec 2025

New paper from Shreya Sinha Roy, me, Christian Robert and Rito Dutta on "Prequential posteriors". We use a a predictive-sequential loss function to fit deep generative forecasting models for the purpose of data assimilation, with waste-free SMC for fitting the neural network arxiv.org/abs/2511.1...

17 Nov 2025

Warwick Statistics is recruiting two Assistant Professors! We are looking for a researcher with a strong track record and outstanding promise in one or more areas of Statistical Methodology and Theory, Computational Statistics and Machine Learning. Closing date: 25th Jan 26 warwick-careers.tal....

Richard Everitt reposted