Dayi (David) Li

Ph.D. Candidate at the University of Toronto

CANSSI Ontario Multidisciplinary Doctoral (Mdoc)

Data Sciences Institute (DSI) Doctoral Fellow


I am a Ph.D. Candidate in statistics advised by Gwen Eadie, Patrick Brown, and Roberto Abraham at the University of Toronto. I am a CANSSI Ontario Multidisciplinary Doctoral (Mdoc) trainee and my research is supported by the Data Sciences Institute Doctoral Fellowship. My research interests lie in the emerging new field of astrostatistics as well as spatial statistics and Bayesian computation. My Ph.D. work focuses on developing novel methodology in spatial statistcs to better understand the mysterious ultra-diffuse galaxies.

Before coming to UofT, I obtained my bachelor degree in financial modelling in 2018 at Western University. Subsequently, I obtained my Masters degree in statistics in 2020 at Western University under the supervision of Ian McLeod and Pauline Barmby.

In general, I believe there is still an inadequacy of rigorous applications of advanced statistical methods in astronomy. Filling this inadequacy can bring about new understanding and discovery in astronomy and help us better understand the Universe. Hence, I am interested in novel statistical application in astronomy as well as developing new statistical methods to tackle challenging astronomical questions.


  • Astrostatisics
  • Spatial Statistics
  • Bayesian Inference and Computation
  • Statistical Pattern Recognition


  • Ph.D., Statistics, 2025 (expected)

    University of Toronto

  • M.Sc., Statistics, 2020

    Western University

  • Honors B.Sc., Financial Modelling, 2018

    Western University


Coming soon…

Fast Inference using Spline Approximation for Detecting Ultra-Diffuse Galaxies by Globular Clusters while Accounting for Uncertainty on Globular Clusters with Data Augmentation

Desiging fast inference algorithm for reversible-jump type MCMC algorithm to facilitate UDG detection in massive astronomical survey.

A Principled Bayesian Estimation of the Globular Cluster counts in Ultra-Diffuse Galaxies using thinned Poisson Point Processes

Constructing a principled Bayesian hierarchical point process model to accurately and efficiently infer the globular cluster counts in …

Recent & Upcoming Talks


Poisson Cluster Process Models for Detecting Ultra-Diffuse Galaxies

We propose a novel set of Poisson Cluster Process models to detect Ultra-Diffuse Galaxies (UDGs), a recently-discovered class of …

Gibbs Point Process Model for Young Star Clusters in M33

We demonstrate the power of Gibbs point process models from the spatial statistics literature when applied to studies of resolved …