Colloquium Speaker:
Dr. Mai Ghannam (PhD-Stats, Univerity of Windsor Alumna)
Postdoctoral Fellow, University of Toronto
Title: Block Maxima Method Under Heteroskedasticity
Abstract: Extreme value theory (EVT) is used to model the risk of extreme, rare events across fields such as climate, finance, and engineering. While recent work has examined the effect of heteroskedasticity in the Peaks-over-Threshold (PoT) framework, the impact on Block Maxima (BM) method remains unexplored. In this talk, I present the first theoretical analysis of BM method under heteroskedasticity. Specifically, I show that the classical BM maximum likelihood estimator for the tail index becomes asymptotically inconsistent when the data exhibit heteroskedasticity, even if the true tail index is constant over time. To address this, I propose two new estimators and establish their consistency and asymptotic normality. Finally, I introduce a hypothesis testing procedure for detecting heteroskedasticity in extreme value data within the BM framework. This is a joint work with Rafał Kulik (University of Ottawa) and Stanislav Volgushev (University of Toronto).
Day & Time: Thursday, November 20, 2025, at 3:00pm
Location: Lambton Tower, Room 9-118