Meet the keynote speakers

Associate Professor Margarita Moreno-Betancur established and leads an internationally recognised hub of innovation and expertise in statistical methods for observational studies, particularly in causal inference. Her integrated research program encompasses the development of innovative methods, collaboration in clinical and public health studies, statistical training for non-statisticians, and biostatistics capacity building.

After a BSc in Mathematics at Universidad de los Andes (Bogota) and an MSc in Statistics at Université Pierre et Marie Curie (Paris), she completed a PhD in Biostatistics in 2014 at Université Paris-Sud (Paris). She held postdoctoral roles in France (Inserm, Paris) and the Victorian Centre for Biostatistics (ViCBiostat, Melbourne), following which she has been competitively self-funded, first with an ARC Discovery Early Career Researcher Award (DECRA, 2019-22) and now an NHMRC Investigator Grant (EL2, 2022-26). In April 2022, she was appointed co-Director of the Clinical Epidemiology and Biostatistics Unit (CEBU) at the Murdoch Children’s Research Institute (MCRI) and The University of Melbourne (UoM).

She conducts methodological research in the areas of causal inference, missing data and survival analysis, to which she and her team have made significant and innovative contributions that are impacting the field of biostatistics and beyond, by influencing the design and analysis of health research studies. She and her team have collaborated on studies in a range of areas, particularly in the life course and social epidemiology, and provide important training and mentoring to non-statisticians. A highlight is an innovative and widely attended short course on methods for observational studies, with a particular focus on cutting-edge causal inference concepts and methods. She has made broad-ranging contributions to supervising, training and mentoring the next generation of Australian biostatisticians, at the Masters, PhD and postdoctoral levels. She is part of the steering groups of ViCBiostat, the MCRI's LifeCourse Initiative and the Master of Biostatistics at UoM.

Title: Causal machine learning in health data science

Abstract: The ultimate goal of medical and health research is to improve patient outcomes and population health. As a result, the overwhelming majority of clinical and public health research studies ask “causal” questions, concerning the effect of treatments, policies, behaviours and other exposures on health outcomes. In many cases, especially in the current era of data deluge, these studies rely on observational (non-experimental) data to address causal questions. Unfortunately, for a long time the statistics discipline largely shunned the possibility of causal inference beyond randomised trials, and instead focused on the development of tools such as regression models without clarity regarding their usefulness and limitations for addressing the causal questions that substantive areas continued to ask. In recent decades, however, the discipline has seen the rise of a new area focused on determining the settings and approaches that could allow causal inference from observational data. This talk will first provide an overview of some of the fundamental contributions of this statistical area to enable and improve the study of causality in health research, and then describe the role of machine learning within this causal inference paradigm, including recent methodological advances.

Marijke’s passion is connecting data and people. She is a Research Fellow at the Australian Red Cross Lifeblood. With over 10 years of experience in statistical consulting in academia, government, and industry, she prides herself on connecting with collaborators and use her skills to enhance the story behind their data. Throughout her career, she has been an advocate for good statistical practice and the importance of telling the story behind the numbers with confidence and impact. As an applied statistician, she is not afraid to get her hands “dirty” to gain more understanding of the data generation process.

Title: The tale of a career with large variance and a central tendency of promoting good statistical practice

Abstract: Starting out studying a degree in psychology, my career has taken many unexpected turns. Whilst it would be easy to state that this was always part of the plan, the truth is more like a random walk being the driver. I have counted moths and mice. I have been hooked up to machines. I have received electroshocks. And that all for the greater good of statistics.

In this lecture I will walk you through the varied path of my career, lessons I learned along the way and demonstrating that the career of an applied statistician can have large variance, but at the centre of it is the quest for improving statistical practice in all application areas.

Andrew Zammit-Mangion completed his PhD at the University of Sheffield, UK, in 2012, and is currently an Associate Professor at the School of Mathematics and Applied Statistics at the University of Wollongong, Australia. His key interests lie in spatial and spatio-temporal models and the inferential tools that enable their use in environmental applications. He has published extensively on several topics in this area, including dynamical spatio-temporal modelling; spatio-temporal point process modelling; Bayesian hierarchical modelling; multivariate spatial modelling; and approximate Bayesian inference. He has also released software packages that facilitate spatial and spatio-temporal modelling. Andrew has worked extensively in two subfields of the geosciences: (i) Antarctic processes and their contribution to sea-level rise, and (ii) trace gases and their transport in the atmosphere. His work in (ii) has contributed to a framework for estimating carbon dioxide sources and sinks from satellite data (coined WOMBAT – the WOllongong Methodology for Bayesian Assimilation of Trace-gases) that is the recipient of a 2023 prize by the Section of Statistics in Physical Engineering Sciences (SPES) of the American Statistical Association. Recently, Andrew has also been looking at novel ways to combine deep learning models with spatio-temporal models, and at ways to use AI to accelerate inference with these models.

Andrew was awarded the Cozzarelli Prize Class III (Best PNAS paper in Engineering and Applied Sciences) by the National Academy of Sciences of the US in 2013, a Discovery Early Career Research Award by the Australian Research Council in 2017, the Abdel El-Shaarawi Young Researcher's Award by The International Environmetrics Society in 2020, and the Early Investigator Award by the Section of Statistics and the Environment of the American Statistical Association in 2022. He is currently a member of NASA's Orbiting Carbon Observatory-2 (OCO-2) Science Team, and in 2019 he published a co-authored book with Distinguished Professors Christopher Wikle and Noel Cressie on spatio-temporal modelling with R.

Title: Fast statistical inference with neural networks and amortisation: Golden ticket or red herring?

Abstract: Neural networks can provide solutions to tasks that were inconceivable just a few years ago and have benefitted society in numerous ways. These benefits primarily stem from a property often referred to as "amortisation": Training a neural network usually requires significant effort and resources but, once trained, the network can solve similar problems repeatedly and rapidly with virtually no additional computational cost. Hence, the substantial initial training cost of training neural networks is "amortised" over time. Amortisation can also be used to enable fast inference with parametric statistical models: Once a network is trained using observational data as input and inferential statements (e.g., point parameter estimates) as output, the network can make inference with future data in a tiny fraction of the computing time needed by conventional likelihood or Monte Carlo methods. These amortised inferential tools have several compelling advantages over classical methods: They do not require knowledge of the likelihood function, are relatively easy to implement, and facilitate inference at a substantially reduced computational cost. In this lecture I will first give a brief review of recent work that has leveraged the property of "amortisation" in statistical inference. I will then evaluate the merits and drawbacks of amortised inference from a statistician's perspective and conclude by outlining the challenges that need to be overcome for these inferential tools to gain widespread acceptance.