Online Program Home
My Program

Abstract Details

Activity Number: 409 - Bayesian Space-Time Modeling
Type: Contributed
Date/Time: Tuesday, July 30, 2019 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #304343 Presentation
Title: Bayesian Nested Lasso with Application to Mixed Frequency Data
Author(s): Satyajit Ghosh* and Kshitij Khare and George Michailidis
Companies: Rutgers University and University of Florida and University of Florida
Keywords: Time Series; Mixed Frequency Data; Nested Lasso; Nowcasting; Spike and Slab Prior

We propose a fully Bayesian method to jointly estimate lag length and the coefficients in the context of mixed frequency regression, another popular time series model. The key technique in order to select the correct lag is our newly invented prior which we call Bayesian nested lasso prior. Through a simulated data we have established that this prior performs better than existing model/lag selection methods when the covariates have a natural ordering. Another key aspect of the Bayesian nested lasso prior is that it can incorporate decaying nature of the signal strength with respect to lag length in context of time series regression. By introducing a logistic functional form in the prior variance we can even reduce the effective dimension of the linear model which is essential to forecast a low frequency variable based on available high frequency covariates.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program