Date: Friday, January 31, 2025
Time: 12:00–1:00 p.m. Atlantic time
Location: University of New Brunswick (Saint John), room to be confirmed, and on Zoom
This talk will be presented by Dylan Spicker, an Assistant Professor in the Department of Mathematics and Statistics at the University of New Brunswick (Saint John). It’s the fourth event in the Atlantic Canada Data Science Tour, a hybrid seminar series organized by CANSSI Atlantic and geared toward upper-level undergraduates in statistics or computer science programs. The host will be Joanna Mills Flemming, Professor in the Department of Mathematics and Statistics and Associate Dean of Graduate and Global Relations at Dalhousie University. Joanna is also the Regional Director of CANSSI Atlantic.
We invite you to join us in person or online! (We’ll send you the Zoom link when you register.)
REGISTER ON EVENTBRITE (opening soon)
The collection and analysis of data have become ubiquitous across nearly every domain (including healthcare, social media, and government). Much of the information that is collected, stored, and analyzed is private or sensitive, and as such, there is increasing pressure to ensure that individual privacy is maintained. Failure to do so can have serious consequences for the individuals involved. Unfortunately, privacy researchers have demonstrated that every statistical analysis can inadvertently leak private information unless the analysis is designed to meet rigorous standards of privacy. This is true even when the data have been “anonymized” by removing personal identifiers (such as names, addresses, or social insurance numbers). In this talk, we will explore these privacy pitfalls and outline the work that researchers are doing to overcome them.
Dylan Spicker (they/them) is an Assistant Professor at the University of New Brunswick (Saint John).
They completed their PhD at the University of Waterloo in the summer of 2022 and their postdoc at McGill University in 2023.
Dylan’s research focuses on areas of causal inference, and specifically methodologies related to dynamic treatment regimes. During their graduate studies, their research focused on measurement error and causal inference. Briefly, measurement error occurs whenever we are interested in measuring something and we do a bad job of it. This happens in almost every study that is run, and unfortunately means that the conclusions that we draw may not be accurate; statistical work on measurement error tries to correct this. Causal inference asks questions of the form “Does X cause Y?” (For instance, “Does smoking cause lung cancer?” (Yes, it does.)) They have a keen interest in providing a theoretical basis for (comparatively) straightforward methods, which are easy to use for non-statisticians, while exhibiting provably good theoretical properties.
During their postdoc, Dylan explored problems related to privacy and dynamic treatment regimes, where they sought to determine ways that an individual’s personal health data can be protected, while gleaning the useful insights that we seek.
Outside of causal inference and measurement error, Dylan is interested in machine learning, and in particular in trying to establish a statistical basis for novel machine learning techniques (including questions related to inference, interpretability, and model selection).
Dylan previously did an undergraduate degree in Finance and Mathematics at Queen’s University (they transferred there after completing their first year at the University of Waterloo/Wilfrid Laurier University in the “Double Degree” program), and a Master of Statistics at Waterloo.
Outside of their research, they pay very close attention to sports, mostly hockey (and how statistics is, or should be, applied there), play music (without any connection to statistics), and enjoy board/video games (with varying degrees of statistical relevance). They have a cat (Charles) who is wonderful.