Topics on Bayesian learning

Brazilian Meeting of Econometrics

Brazilian Econometric Society

December 2020

Hedibert Freitas Lopes

Professor of Statistics and Econometrics

Head of the Center of Statistics, Data and Decision Sciences

Insper Institute of Education and Research

http://www.hedibert.org

Part 1. Bayesian ingredients

Brief overview of key Bayesian ingredients and computation

· Physicists A, B, C and D

· Banana-shaped posterior

· Model selection/comparison: Bernoulli, logit/probit regression?

· Boston housing data

Part 2. Linear models

Variable selection and regularization

The illusion of the illusion of sparsity

Part 3. Time-varying variance

GARCH modeling

Stochastic volatility modeling

· Petrobras returns: GARCH(1,1) vs SV-AR(1) with t-errors

Factor stochastic volatility modeling

Dynamic sparsity in dynamic regressions