Bayesian Inference, Metropolis-Langevin and Hamiltonian in Riemann manifolds.

• Grant number: 15/00627-9
• Support Opportunities: Scholarships abroad - Research
• Start date until: August 01, 2015
• End date until: July 31, 2016
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Applied Probability and Statistics
• Principal Investigator: Ricardo Sandes Ehlers
• Grantee: Ricardo Sandes Ehlers
• Host Investigator: Nial Friel
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Institution abroad: University College Dublin, Ireland

Abstract

In this project, recently proposed computationally intensive methods which explore ideas of Riemann geometry in the parameter space will be studied and applied in statistical models under the Bayesian approach. These methods seek to adapt to the local structure of a target distribution which can potentially turn Markov chain Monte Carlo methods (MCMC) more efficient. The computational costs involved may be high and therefore the cost-benefit must be investigated through simulation studies and real data analyses.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

ZEVALLOS, MAURICIO; GASCO, LORETTA; EHLERS, RICARDO. Riemann manifold Langevin methods on stochastic volatility estimation. COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, v. 46, n. 10, p. 7942-7956, 2017. (15/00627-9, 13/00506-1)

HARTMANN, MARCELO; EHLERS, RICARDO S.. Bayesian inference for generalized extreme value distributions via Hamiltonian Monte Carlo. COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, v. 46, n. 7, p. 5285-5302, 2017. (15/00627-9)