A special , interdisciplinary, international  Symposium on Causality  has been organized by the Department of Statistics, Computer Sciences and Applications of the University of Florence in collaboration with the Harvard Data Science Initiative, the European University Institute, and other stakeholders.

The Symposium on Causality  was hosted at the Museo Galileo.

The event was one of several planned to celebrate the centenary of the University of Florence.

The Symposium has been a collection of contributions on the state of the art and future challenges in causality and causal inference for the new era of Data Science.

Four keynote lectures (Guido Imbens, Elizabeth Stuart, Mihaela van der Schaar, Alessandro De Angelis) on the field’s past, present and future from different perspectives and disciplines, plus four roundtables on hot topics in causality and causal inference with experts from academia and the industry have been planned.

Program

List of Participants

Photos

Video

Keynote Speakers

Guido Imbens

Applied Econometrics Professor and Professor of Economics, Graduate School of Business, Stanford University

Causal Inference with Panel Data: New Methods and Remaining Challenges

In recent years much work has been done on methods for credibly estimating causal effects in settings with panel data. These methods, including synthetic control methods, matrix completion methods and variations thereon have have greatly improved our understanding of the challenges in estimating causal effects with longitudinal data. In this presentation I will describe some of these insights, and also discuss shortcomings of the existing methods. In particular these include the lack of accounting for the time series aspect of the longitudinal data and the challenges for inference. I will discuss new estimators that take the time series dimension more seriously and show how they improve estimation. I will also discuss new methods for inference that improve the power of testing for the presence of causal effects. (based on joint work with Alex Almeida,  Susan Athey, Alexia Olaizola,  Zhaonan Qu, Alex Almeida,  Eva Lestant, and Davide Viviano).

Elizabeth A. Stuart

Frank Hurley and Catherine Dorrier Professor of Biostatistics, Johns Hopkins Bloomberg School of Public Health

Integrating data for causal inference

Many causal questions of interest cannot be answered through analysis of a single dataset, and as data becomes increasingly available, there is more and more interest in leveraging that data to answer nuanced questions. Such questions might include examining the generalizability of randomized trial results to target populations, to better understanding of effect heterogeneity by combining small (unbiased) randomized trials with large (but confounded) non-experimental data sources. This talk will discuss methods for causal inference in such integrated datasets, including both the promise and potential for doing so, as well as implementation challenges, such as when the measures in the different data sources are discordant. Motivating examples will come from medicine and public health, and with lessons for a range of fields, and with final comments on the broader field of evidence synthesis for causal inference.

Alessandro De Angelis

Full Professor of Experimental Physics, University of Padova and University of Lisboa

Causality in physics: from Galileo to Einstein

Galileo’s work established a formal concept of causality in physics by using mathematical descriptions of physical processes to predict the evolution of systems from causes. Building on Galileo’s ideas, Newton formalized this concept within the framework of classical mechanics. In the 19th century, Faraday introduced the concept of field, a region of space where a force operates. Maxwell’s equations formalized this idea, describing how electric and magnetic fields propagate through space, leading to a new understanding of causality. Finally, Einstein revolutionized the concept of causality with his theories of relativity.

Mihaela van der Schaar

John Humphrey Plummer Professor of Machine Learning, Artificial Intelligence and Medicine at the University of Cambridge

The Causal Discovery Ladder: Unravelling Governing Equations and Beyond using Machine Learning

While recent strides in machine learning and statistics have advanced causal inference, several frontiers await exploration in order to make real-world impact. First, I will show how unraveling governing equations from temporal trajectories can deepen our understanding of dynamical systems through causal inference over time. Second, I will explore how causal deep learning can help close the gap between theory and practice, providing practitioners with powerful new ways of thinking and tools. Finally, I will discuss how leveraging real-world observational data can transform various aspects of clinical trials. By pushing these boundaries, we aim to bring causality into practical use, unlocking insights with transformative potential in fields like medicine.

Organizing Committee

L'articolo Symposium on Causality proviene da Rethinking Data Science.

It was an exciting free 4-day meeting where young scholars were prompted to explore statistics and data science to extract new knowledge from data. During the camp, the young scholar participants worked in teams under the guidance of five senior researchers Bianca De Stavola, Daniele Durante, Laura Forastiere, Antonio Gasparrini and Corwin Zigler with leading experience in different areas of Statistics, including  statistical methods for long-term longitudinal studies, Bayesian methods, models for complex, high–dimensional and network–related phenomena, causal inference in complex settings involving spatial and network data, methods for investigating causal pathways through intermediate variables or mechanisms, development of study designs and statistical for  time series data, quasi-experimental studies, and survival analysis. 

The senior researchers worked hard with the young scholars to develop innovative methods and models to analyze a common dataset, to answer scientific questions arising from a cancer genomics case study. The publicly available data of lower grade glioma from the Cancer Genome Atlas (TCGA) data portal was used as motivating application study. Francesco Stingo, who previously worked on these data, was invited to introduce them on the first day of the Data Research Camp. After the presentation of the dataset and Q&A session on the data, the groups started to work! Innovative ideas, new questions and cutting-age statistical methods were proposed to analyze the data as shown by the mid and final presentations. 

The DiSIA is proud to have organized this exciting event; it laid the groundwork for new interdisciplinary pathways that can lead to cutting-age contributions to the modern Statistics from both methodological and applied perspectives. 

The Data Research Camp was supported by the excellence program “Rethinking Data Science (ReDS): advances in causal inference and statistical learning across domains” and it is endorsed by the PhD programs in Statistics of the University of Florence, Bocconi University, University of Bologna and University of Padova. It followed two successful editions organized by the Department of Statistical Sciences of the University of Padua (see the websites of the 2019 and 2022 editions).

Programme

Participants

Photos

---

Senior Group Leaders

Bianca De Stavola

UCL University College London, UK

Daniele Durante

Bocconi University, Italy

Laura Forastiere

Yale University, USA

Antonio Gasparrini

LSHTM – London School of Hygiene & Tropical Medicine

Corwin Zigler

The University of Texas at Austin, USA

L'articolo The Data Research Camp of DiSIA 2024 proviene da Rethinking Data Science.