SRE 2025
The following seminar series of Tuesday Talks will be held as part of the 2025 Warwick Statistics Summer Research Experience organized by Richard Everitt, Brett Kolesnik and Matthew Thorpe.
All talks will be held on Tuesdays 12–1pm in MB2.22.
Everyone is welcome.
A catered lunch will be provided for summer research students in the Statistics Common Room (1st floor in MSB) following the talks during 1–2pm.
Tuesday, July 8
*~*~* Janique Krasnowska *~*~*
12–12:25pm (MB2.22)
Genealogy of Branching Processes
Branching processes are used to model a population where the population size varies non-deterministically. A fundamental question in statistical genetics is about the time until the most recent common ancestor of a sample of individuals at the present time. In this talk, we are going to introduce the most important concepts in branching processes and explore how they influence the genealogical tree.
*~*~* Gregor Steiner *~*~*
12:30–12:55pm (MB2.22)
Bayesian Model Averaging in Causal Instrumental Variable Models
I will briefly introduce instrumental variable regression, a popular method to infer causal effects under unobserved confounding, and present our recent work on Bayesian model averaging in instrumental variable models. I will demonstrate the usefulness of our method in an empirical application estimating the causal effect of education on income.
Tuesday, July 15
*~*~* Kieran Drury *~*~*
12–12:25pm (MB2.22)
Elicitation of Bayesian Networks through Expert Judgement
Bayesian networks are a powerful tool for intuitively modelling complex environments – hence being of particular benefit when used to support a client’s decision making. However, the vast number of dependencies that Bayesian networks feature often mean that any available data is insufficient for reliably determining each parameter in each of the conditional probability tables across the network. Instead, we can utilise expert judgement to inform us of these parameters, as well as for guiding the choice of network structure. This talk introduces how expert judgement can be carefully elicited to construct a Bayesian network that is faithful to true expert beliefs about a complex real-world system. We will further discuss how such a network can then be used to guide decision making with an application to governmental policy choice for supporting pollinator abundance in Australia.
*~*~* Salvatore Ciano *~*~*
12:30–12:55pm (MB2.22)
Stochastic decision models in behavioural economics: optimal stopping under Prospect Theory preferences
There is a well-known intuition linking Prospect Theory (PT) (A. Tversky, D. Kahneman, 1992) and the disposition effect, the tendency to sell assets that have increased in value while holding on to assets that have declined. Several optimal stopping models have been proposed to capture this phenomenon; this talk will mostly focus on the seminal work presented in "Prospect Theory, Liquidation, and the Disposition Effect" (V. Henderson, 2012), where PT preferences are imposed through an S-shaped utility function.
Tuesday, July 22
*~*~* Pablo Ramses Alonso Martin *~*~*
12–12:25pm (MB2.22)
Slow-Fast Differential Equations and Statistical Inference
Multiscale models describe evolving systems where dynamics unfold on multiple interacting time or spatial scales, common in physics, biology, and finance. This talk provides an accessible overview of multiscale (stochastic) differential equations, highlighting the key concepts of homogenization and averaging. We also address the problem of statistical inference in these models, including challenges in parameter estimation and techniques that leverage the multiscale structure for improved accuracy.
*~*~* Alastair Crossley *~*~*
12:30–12:55pm (MB2.22)
Stochastic Modelling of Electron-Photon Cascades via Branching SDEs
Electron-photon cascades are central to high-energy physics, with applications in astrophysics and radiation therapy. This project develops a new stochastic framework using branching processes and stochastic differential equations to model these cascades pathwise, even in regimes with infinite interaction rates. The approach offers a probabilistic extension of classical models.
Tuesday, July 29
*~*~* Usman Ladan *~*~*
12–12:25pm (MB2.22)
An introduction to branching processes (and their approximations)
In this talk, we introduce branching processes and discuss the behaviour of both discrete and continuous time models. These objects have applications in a wide variety of areas such as biology and physics. If we have time, we will also discuss how to efficiently approximate these objects.
*~*~* Federico Perlino *~*~*
12:30–12:55pm (MB2.22)
Inference from Deep Gaussian Processes over Directed Acyclic Graphs
Deep Gaussian Processes (DGPs) extend Gaussian Processes by modeling data through hierarchical compositional structures, yet most existing approaches are confined to sequential (chain-like) architectures. This presentation introduces a variational inference framework for DGPs defined over Directed Acyclic Graphs, allowing observations at intermediate nodes and enabling more expressive models. This work lays the inferential groundwork for integrating diverse inductive biases—from deterministic laws to causal relationships—within a unified probabilistic framework.
Tuesday, August 5
*~*~* Geyu Ji *~*~*
12–12:25pm (MB2.22)
The Analysis for Deuteriation Process with MCMC
This project aims to describe and forecast the deuteriation process of a chemical. The reaction is simplified as a Ferris Wheel model and considered as a Markov Process. Random-walk MCMC is applied to sample the posterior distributions of unknown variables given the real observation data.
*~*~* Florian Gutekunst *~*~*
12:30–12:55pm (MB2.22)
Stochastic Control and the Merton Problem
We introduce the theory of stochastic control, which deals with the question of how to best set the dynamics of a random process to maximise some reward. The theory is illustrated with Merton’s investment and consumption problem. There, an investor aims to maximize their utility from consumption, while choosing how much money to invest in a stock, keep in a bank account, or consume.
Tuesday, August 12
*~*~* Edwin Tang *~*~*
12–12:25pm (MB2.22)
The Trouble with the Sample Mean (and How to Fix It)
We’re all taught early on to use the sample mean to estimate the true mean of a distribution—but is that always the best choice? In this talk, we’ll explore situations where the sample mean works well, and where it starts to break down. Using the powerful toolkit of concentration inequalities (you may already know Markov’s inequality), we’ll discuss why the alternative estimators, such as the median and the median of means, can be more reliable.
*~*~* Vinayak Niraj *~*~*
12:30–12:55pm (MB2.22)
Generative AI, SDEs and Convergence
How do the latest image generation models like Dall-E work? Why are they called Diffusion models? What is their connection with Stochastic Differential Equations (SDEs)? Why do we care about these SDEs' convergence rates, and how do we find them?
Tuesday, August 19
*~*~* Peter Matthews *~*~*
12–12:25pm (MB2.22)
An invitation to the Dirichlet Process
Imagine picking out a ball from an urn, then when you put it back you add another ball of the same colour and repeat the process. You want to describe the probability of seeing a particular sequence of colours, and see what properties of the sequence you expect to hold. Now imagine that you want to estimate the density of some data that you see, but you want to be fully Bayesian about it. Amazingly, the same beautiful mathematical object, the Dirichlet Process, underlies both settings. I hope to give you a taste of why it is so interesting to, and studied by, both pure probabilists and applied statistical modellers.
*~*~* Isabella Goncalves de Alvarenga *~*~*
12:30–12:55pm (MB2.22)
The Multitype Contact Process
The multitype contact process is a mathematical model for population dynamics where two competing species share the same environment. In this talk, we will define the model and explore key features of the region where the two species interact, particularly along their shared boundary.
Tuesday, August 26
*~*~* Prof. Anastasia Papavasiliou *~*~*
12–1pm (D0.06)
Warwick Statistics CDT Program Info Session with Q&A
The director of the Warwick Statistics Centre for Doctoral Training (CDT) will give a presentation on the CDT program at Warwick Statistics. There will be time to answer any questions you have about the program.