Statistics Seminars are held on Wednesdays 14:00 – 15:00. Everyone is welcome! We gather for coffee/tea and biscuits around 15 minutes before the seminar begins.
Statistics seminar dates alternate with the CREEM seminar series. Some seminars are joint with CREEM.
Some of the seminars this year will be held in-person and some online. The in-person seminars will be held in either the Mathematical Institute or the Observatory seminar room. Please see below for more details.
Forthcoming statistics seminars
16-Nov, in-person talk, Dr Pantelis Samartsidis, Investigator Statistician, MRC Biostatistics Unit, University of Cambridge
Title: A Bayesian multivariate factor analysis model for causal inference using time-series observational data on mixed outcomes
Abstract: Assessing the impact of an intervention by using time-series observational data on multiple units and outcomes is a frequent problem in many fields of scientific research. Here, we propose a novel Bayesian multivariate factor analysis model for estimating intervention effects in such settings and develop an efficient Markov chain Monte Carlo algorithm to sample from the high-dimensional and non-tractable posterior of interest. The proposed method is one of the few that can simultaneously deal with outcomes of mixed type (continuous, binomial, count), increase efficiency in the estimates of the causal effects by jointly modelling multiple outcomes affected by the intervention (as shown via a simulation study), and easily provide uncertainty quantification for all causal estimands of interest. Using the proposed approach, we evaluate the impact that Local Tracing Partnerships had on the effectiveness of England’s Test and Trace programme for COVID-19.
30-Nov, joint talk with CREEM, Dr Blanca Sarzo, University of Valencia
2022-23 Semester 1
14-Sept, Online talk, Dr Dennis Prangle, Senior Lecturer in Statistics, School of Mathematics, University of Bristol
We approximate the posterior using normalizing flows, a flexible parametric family of densities. Training data is generated by ABC importance sampling with a large bandwidth parameter. This is “distilled” by using it to train the normalising flow parameters. The process is iterated, using the updated flow as the importance sampling proposal, and slowly reducing the ABC bandwidth until a proposal is generated for a good approximation to the posterior. Unlike most other likelihood-free methods, we avoid the need to reduce data to low dimensional summary statistics, and hence can achieve more accurate results.
28-Sept, In-person talk, Rachel Phillip, Medical Statistician, Clinical Trials Research Unit, University of Leeds
Rachel is an alumnus of our School and the talk should be of interest to both students and staff.
Title: Working as a statistician on Phase I cancer clinical trials
Abstract: Clinical trials are research studies that are conducted in people in order to study and test new medical treatments. Trials are usually conducted in phases that build on each other, with Phase I trials being the first steps of testing new treatments in people. There is often limited safety information on new treatments, so the primary aims of Phase I studies are to ascertain the safety profile of the intervention and to determine the highest dose that can be given safely without severe side effects that can be taken forward for further investigation in future studies. This talk will provide an introduction to the different areas that a statistician works on in clinical trials, the common statistical designs of Phase I studies as well as talking about CONCORDE – an innovative phase I platform trial testing different drug-radiotherapy combinations.
05-Oct, Online talk (attending from Maths Tutorial Room 1A), Prof Alexandros Beskos, Professor in Statistics, UCL
Abstract: Bayesian inference for nonlinear diffusions, observed at discrete times, is a challenging task that has prompted the development of a number of algorithms, mainly within the computational statistics community. We propose a new direction, and accompanying methodology, borrowing ideas from statistical physics and computational chemistry, for inferring the posterior distribution of latent diffusion paths and model parameters, given observations of the process. Joint configurations of the underlying process noise and of parameters, mapping onto diffusion paths consistent with observations, form an implicitly defined manifold. Then, by making use of a constrained Hamiltonian Monte Carlo algorithm on the embedded manifold, we are able to perform computationally efficient inference for a class of discretely observed diffusion models. Critically, in contrast with other approaches proposed in the literature, our methodology is highly automated, requiring minimal user intervention and applying alike in a range of settings, including: elliptic or hypo-elliptic systems; observations with or without noise; linear or non-linear observation operators. Exploiting Markovianity, we propose a variant of the method with complexity that scales linearly in the resolution of path discretisation and the number of observation times. The talk is based on a forthcoming JRSSB paper: https://doi.org/10.1111/rssb.12497
19-Oct, In-person talk joint with CREEM, Dr Ben Swallow, Lecturer in Statistics, University of St Andrews
Title: Bayesian causal inference for zero-inflated GLMs using a potential outcomes framework’
Abstract: We propose a method for conducting Bayesian causal inference under a generalised linear model potential outcomes framework, for data where there are many more zeros than would naturally be expected. We develop an approach using both semi-continuous and fully continuous probability distributions and apply the approach to both simulated data and ornithological citizen science data in the UK, comparing the results to purely observational studies. Further analyses of the contrasting GLMs are also discussed.
26-Oct, JJ Valletta Memorial lecture: Dr TJ McKinley, Lecturer in Mathematical Biology, Department of Mathematics and Statistics, University of Exeter. In person in Lecture Theatre D, Mathematical Institute.
Title: Emulation-driven inference for complex spatial meta-population models
Abstract: Calibration of complex stochastic infectious disease models is challenging. These often have high-dimensional input spaces, with the models exhibiting complex, non-linear dynamics. Coupled with this is a paucity of necessary data, resulting in a large number of hidden states that must be handled by the inference routine. Likelihood-based approaches to this missing data problem are very flexible, but challenging to scale due to having to monitor and update these hidden states. Methods based on simulating the hidden states directly from the model-of-interest have the advantage that they are often much more straightforward to code, and thus are easier to implement and adapt to changing model structures. However, they often require very large numbers of simulations in order to adequately explore the input space, which can render them infeasible for many large-scale problems.
This seminar will be given in the memory of our colleague JJ Valetta, who suddenly passed away while hillwalking in October 2020. The seminar will be followed by a reception in the Mathematical Institute Common Room (on the ground floor). All are welcome, both to the lecture and the reception!
02-Nov, joint with CREEM, Dr Wei Zhang, Lecturer in Statistics, School of Mathematics and Statistics, University of Glasgow
09-Nov, online talk, Prof Chris Holmes, Professor in Biostatistics at the Departments of Statistics and the Nuffield Department of Medicine, University of Oxford
Title: Bayesian Predictive inference
Abstract: De Finetti promoted the importance of predictive models for observables as the basis for Bayesian inference. The assumption of exchangeability, implying aspects of symmetry in the predictive model, motivates the usual likelihood-prior construction and with it the traditional learning approach involving a prior to posterior update using Bayes’ rule. We discuss an alternative approach, treating Bayesian inference as a missing data problem for observables not yet obtained from the population needed to estimate a parameter precisely or make a decision correctly. This motivates the direct use of predictive models for inference, relaxing exchangeability to start modelling from the data in hand (with or without a prior). Martingales play a key role in the construction. This is joint work with Stephen Walker and Edwin Fong, based on the paper “Martingale Posteriors” to appear with discussion JRSS Series B.