Groupe d’études et de recherche en analyse des décisions

Simulation and CDF estimation for truncated multivariate normal and student-\(t\) distributions

Zdravko I. Botev The University of New South Wales, Australie

We propose an exponential tilting method for the accurate evaluation of truncated the multivariate normal and student-\(t\) distributions. The approach is also effective for simulating from these distributions in high dimensions. Numerical and theoretical results show that the method improves significantly on the current best methodology. Applications of the method include exact sampling from the Bayesian posterior of the probit regression model.