Stat Colloquium: Dr. Reetam Majumder
University of Arkansas
Location
Mathematics/Psychology : 104
Date & Time
October 31, 2025, 11:00 am – 12:00 pm
Description
Title: Vecchia approximated density regression for spatial models with intractable likelihoods
Abstract: Extreme value analysis is critical for understanding the effects of climate change. Exploiting the spatiotemporal structure of climate data can improve estimates by borrowing strength across nearby locations and provide estimates of the probability of simultaneous extreme events. The max-stable process is the cornerstone of several models for spatially-dependent extremes. While the process is theoretically justified, it leads to an intractable likelihood function that makes exact inference unviable for practical problems. We propose to use deep learning to overcome this computational challenge. A surrogate likelihood, comprising a product of univariate conditional likelihoods, is formulated using the Vecchia approximation. We generate millions of draws from the underlying data-generating process under different parameter configurations and then use a deep density regression method to learn a mapping from the data and parameters to the surrogate likelihood. We verify through extensive simulation experiments that the surrogate likelihood leads to reliable Bayesian inference for the process parameters, and present a case study using extreme streamflow data for the contiguous United States.