Edward Gunning (University College Cork, Ireland) 

Title: Generative Regression Models for Functional and Object Data 

 Abstract: Classical multivariate statistical approaches often view each observation as a vector in Euclidean space. However, modern technologies enable the collection of data in which observations present as more general “objects", e.g., curves, images, shapes, and networks. Functional data analysis (FDA) is a canonical example, where observations are measured over some continuum and are viewed as smooth, continuous functions rather than sequences of discrete measurements. Developments in functional and object data analysis extend classical multivariate techniques, e.g., for prediction, inference, visualisation and dimension reduction, to operate on more general objects. 

In particular, regression methods allow the objects to be used as predictors, outcomes, or both. In this talk, I will present two of my research projects to exemplify the scope of functional and object regression. The first project extends methods for regressing objects on scalar predictors, using random effects to account for dependence structures. Our key contribution is to introduce general non-linear transformations of the objects that can be learned flexibly using modern deep learning methods, e.g., convolutional neural networks, to induce a generative non-linear model for them. In the second project, we develop and extend a framework for regressing a function’s highest-order derivative on its lower-order derivatives, thereby estimating a differential equation that describes the dynamic relations in the data. Again, we take a generative model viewpoint of this approach, and develop bespoke tools for estimation, inference and data generation.

Dr Sinéad Moylett Abstract

Title: The Stats Person in the Room: An Early Career Across Neurodegenerative Disease and Real-World Data

Abstract: For most of my career I have been the statistician embedded in clinical research teams, working on real-world data that were never collected for the question being asked. In this talk I introduce myself by walking through the roles I've held so far. At Cambridge, I used health records to study dementia with Lewy bodies. At KU Leuven, I predicted long-term disability in multiple sclerosis from irregular clinical follow-up. At Canterbury, I compared methods for handling missing data in The Irish Longitudinal Study on Ageing (TILDA). Along the way, I'll reflect on what being "the stats person in the room" involves.

Dr Lars Jermiin Abstract

Title:

A new phylogenetic protocol: dealing with model misspecification and confirmation bias

Abstract:

Molecular phylogenetics is a scientific discipline that relies on the mathematical, statistical, and computational sciences and delivers solutions to the life sciences. It plays a key role in comparative genomics and has an increasingly significant impact on industry, government, public health, and society. Here I argue that a widely used phylogenetic protocol is missing two essential steps, and that this absence allows model misspecification and confirmation bias to unduly influence phylogenetic estimates. Building on this, I outline a range of steps that I have taken to improve the accuracy of phylogenetic estimates.

Speaker Biography

Lars Jermiin a graduate from University of Aarhus, with an MSc in Population Genetics (1989), and from La Trobe University, with a PhD in Molecular Evolution (1994). Following six years as Postdoctoral Research Fellow (at La Trobe University, University of Ottawa, and the Australian National University) and Research Fellow (at the Australian National University, and University of Sydney), he became Lecturer (2001), Senior Lecturer (2004) and Associate Professor (2008) at University of Sydney. In 2002, he co-founded and later led the Sydney University Biological and Informatics Centre (2002-2007). In 2009, he invited to the position of Science Leader at CSIRO’s Division of Entomology, where he led the development of Genome Bioinformatics capability and published highly influential scientific papers. In 2016, he became Honorary Professor at the Australian National University. In 2017, he moved to Ireland, where he is now Assistant Professor at University of Galway’s School of Mathematical and Statistical Sciences. In 2026, he was elected Research Affiliate at the Leibniz Institute for Analysis of Biodiversity Change, Bonn, Germany.

Dr Rafael De Andrade Moral Abstract

Title:

Wildlife Monitoring Is About Counting Needles in a Haystack

Abstract:

Monitoring wildlife is both essential and fundamentally messy. Ecosystems (both natural and human-altered) can support endangered, damaging, and economically valuable species, yet estimating animal abundance remains difficult, because detection is imperfect, data are either very expensive to obtain or very sparse, and a number of assumptions often dominate inference. Wildlife counting methods are like looking for a needle in a haystack with the needle usually hiding from you, or occasionally attacking you… This brings about conceptual, technological, and societal challenges: what should indicators indicate? How many indicators are too many? How relevant is spatial and/or temporal heterogeneity? In this talk, I will frame wildlife monitoring as a statistical problem with escalating levels of difficulty and will show how to estimate animal abundance when we can identify individuals, see them but not identify who’s who, or only infer their presence from indirect evidence. Across these settings, I will show how statistical models can be adapted to accommodate real-world constraints. Real case studies include monitoring deer, collared peccaries, foxes, giant anteaters, and wild boars. I will also illustrate how movement, behaviour, and landscape structure can affect data collection, and showcase a counterintuitive result of a common practice aimed at avoiding bias that ends up introducing more bias in camera trap surveys. Ultimately, I will attempt to substantiate an argument for a pragmatic approach to wildlife monitoring that acknowledges conceptual, technological, and societal constraints, while equipping statisticians to ask better questions of imperfect data in the real world.

Dr Hassan Alkhayuon Abstract

Title:

Zombie fires: seasonal forcing, heat waves, and rate-induced phase-tipping

Abstract:

Zombie fires are an interesting phenomenon in peatlands, where fires disappear from the surface in winter while smouldering underground, only to reappear the following spring with the seasonal rise in atmospheric temperature. In this talk, we present a conceptual model of peat soil combustion that incorporates seasonal variations in atmospheric temperature as well as temperature anomalies due to heat waves. We demonstrate that a heat wave can cause the bioactive soil to transition from its cold base state to one of two metastable states. The first is a hot state, where the temperature of the peat is below combustion but high enough to sustain high levels of microbial bioactivity (or respiration), releasing large amounts of carbon into the atmosphere over a prolonged period of a few years to a decade. The second is a zombie fire state, where the organic content of the peat burns at a very high temperature and for a shorter period of one to a few years. We then combine concepts from geometric singular perturbation theory with a special compactification technique and our understanding of rate-induced phase-tipping from periodic solutions to explain these transitions.