|
Activity Number:
|
541
- Recent Progresses in Bayesian Inference in Large Parameter Spaces: Jayanta K. Ghosh Memorial Session
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, July 31, 2019 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Memorial
|
|
Abstract #300464
|
Presentation
|
|
Title:
|
Extreme Value Analysis with Semiparametric Density Models
|
|
Author(s):
|
Surya Tokdar*
|
|
Companies:
|
Duke University
|
|
Keywords:
|
Extreme;
Heavy tailed;
Semiparametric;
Bayesian;
Tail index;
Density estimation
|
|
Abstract:
|
This work introduces a transformation based model for probability densities that seamlessly conjoins a nonparametric function representing the bulk of the distribution with a parametric representation of the tails. This semiparametric model offers an exact characterization of the tail index. I will talk about methods for Bayesian inference from this model and present asymptotic frequentist guarantees of the resulting estimates. By utilizing information from the entire data set, this method enables accurate and robust tail index estimation and prediction of high quantiles even from limited data. Illustrative applications are presented to forecasting extreme weather outcomes from limited data. I will discuss challenges and possible avenues of extending the model to multivariate heavy tailed density estimation.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2019 program
|
