Seminar Florian Huber: Recent developments in Bayesian VAR modeling

Recent developments in Bayesian VAR modeling

Seminar by Prof. Dr. Florian Huber, Paris Lodron University of Salzburg

In this talk Florian Huber discusses recent methodological advances in Bayesian multivariate time series models. Events such as the Covid-19 pandemic or the Ukraine war have generated substantial outliers in macroeconomic time series. Inclusion of such outliers often has deleterious effects on parameter estimates and predictive accuracy in linear models. The first part of the presentation deals with nonparametric extensions of VARs to capture nonlinear interactions between the endogenous variables and their lags without imposing strong prior restrictions on potential nonlinearities. The second part drills into the forecasting performance of these various extensions using several different datasets.

Please register for this event at event(at)