Colloquium Speaker:
Dr. Cindy Feng, Ph.D.
Dalhousie University, Halifax, Nova Scotia
Title: Copula-Based Modeling of Multivariate Count Data: Dynamic Dependence and Joint Risk in Health Services
Abstract: Modeling dependence between related health outcomes is key to understanding their joint dynamics, particularly during periods of high system demand. However, multivariate count data are often analyzed under independence assumptions, which can obscure important relationships and bias inference. This talk presents two complementary copula-based frameworks for analyzing correlated, overdispersed count data. The first jointly models monthly emergency department visits for mental health and substance use, capturing how their dependence evolves over time. Using emergency department data from Nova Scotia (2017–2023), the model suggests strong positive dependence prior to COVID-19, a temporary reversal during lockdowns, and partial recovery afterward. The second framework focuses on estimating joint and conditional exceedance probabilities, enabling assessment of concurrent high-burden events and outcome-specific risk. Applied to hospitalization data, results show moderate dependence, spatial variation, and meaningful differences from analyses that assume independence. Together, these approaches provide flexible tools for modeling dependence and assessing joint risk in multivariate count data, with important implications for health system planning.
Day & Time: Thursday, April 30, 2026, at 3:00pm
Location: Zoom
Counts toward seminar attendance for MSc and PhD students in Math & Stats.