We announce the following DEC Statistics seminar:

Thursday, October 24 12:30 Meeting room 3-e4-sr03, Bocconi University, Via Roentgen 1, 3rd floor

Annalisa Cadonna (Wirtschaftsuniversität Wien)

Title: Hierarchical shrinkage in time-varying parameter models through the Triple Gamma prior

Abstract: Time varying parameter (TVP) models are widely used in time series analysis, because of their flexibility and ability to capture gradual changes in the evolution dynamics of the model variables. The popularity of TVP state-space models in macroeconomics and finance is based on the evidence that, in most applications, the influence of certain predictors on the outcome variables varies over time. The risk of overfitting in TVP models is a well know issue, which increases with the dimension of the problem at hand, as only a small subset of the predictors might actually have a dynamic influence on the output variables. Hence, in the last decade, there has been a growing need for models and methods able to discriminate between time varying and static parameters in TVP models or, in other words, perform variance selection. We propose a new global local shrinkage prior for shrinkage of variances in TVP models, referred to as the Triple Gamma prior. The Triple Gamma prior extends to variance selection the Normal-Gamma-Gamma prior introduced in Griffin and Brown (2017) in the context of highly structured regression models. The Triple Gamma prior accommodates, as special or limit cases, the Bayesian Lasso, the Double Gamma prior, and the popular Horseshoe prior. Interesting properties of the triple gamma prior are outlined and an efficient Markov Chain Monte Carlo algorithm is developed. An extended simulation study is conducted and the proposed modeling approach is applied to real data, both in a univariate and a multivariate framework. The predictive performance of different shrinkage priors is compared in terms of log predictive density scores.

Kind regards, Giacomo Zanella

The DEC statistics seminars schedule is available at http://www.unibocconi.eu/statseminar

Please note: if you are a guest and you do not have a Bocconi ID Card to access to the Bocconi Buildings, please communicate your participation by sending an email to arianna.colombo@unibocconi.it