Title of the talk: Divergence-Based Estimation on Power-Law Models: Projection Theorems, Sufficiency, and Cramer-Rao Bounds

Date and Time: 17 April 2026, 5:00 P.M (ONLINE)

Abstract: The solution to parameter estimation problems in statistics often depends on the complexity of the underlying family of distributions. A useful perspective is to recast them as the minimization of suitable divergence (distance) functions. In this talk, I discuss such methods in the context of power-law models, including heavy-tailed distributions such as the Student and Cauchy, where classical approaches may perform poorly. I show how these methods lead to tractable estimators and connect naturally with key ideas in statistics, such as sufficiency and Cramer-Rao bounds. I also highlight a surprising phenomenon: for certain models, methods designed to be robust can lose this property and reduce to classical estimators. This illustrates that robustness depends not only on the method, but also on the underlying model structure.

About the Speaker: Dr. Atin Gayen is currently a Postdoctoral Fellow at the Indian Institute of Technology Roorkee. He received his Ph.D. from the Indian Institute of Technology Palakkad in 2023, where his doctoral research focused on divergence-based approaches to robust statistical inference. His research interests include information geometry, divergence-based methods in statistics, and quantum information theory, with an emphasis on the interplay between geometry, robustness, and inference.