Online Program Home
  My Program

Abstract Details

Activity Number: 281 - Advances in Time Series Methodology
Type: Topic Contributed
Date/Time: Tuesday, August 1, 2017 : 8:30 AM to 10:20 AM
Sponsor: Business and Economic Statistics Section
Abstract #324723 View Presentation
Title: Bayesian Estimation of Optimal Differencing Operator in Cointegrated Systems
Author(s): Anindya Roy* and Tucker McElroy
Companies: University of Maryland at Baltimore County and U. S. Census Bureau
Keywords: Causality ; Invertibility ; Vector Autoregression

Vector autoregressive (VAR) models are common multivariate time series models that are flexible enough to model a variety of processes. Often VAR systems are cointegrated with some linear combinations of the time series resulting in pure unit-root processes while other linear combinations resulting in causal autoregressive processes. To analyze such processes users may work with the differenced time series. However, differencing all series equally results in some linear combinations that are over-differenced and hence non-invertible processes. Non-invertible processes are problematic to deal in terms of long-term forecasting. We use a factorization of the VAR operator along with priors that are constrained to the factor space to estimate an optimal differencing operator that when applied to the original series produces invertible processes.

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

Back to the full JSM 2017 program

Copyright © American Statistical Association