Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 194 - Time Series in Federal Statistics
Type: Topic Contributed
Date/Time: Monday, August 8, 2022 : 2:00 PM to 3:50 PM
Sponsor: Government Statistics Section
Abstract #323100
Title: An Empirical Look at Multivariate Signal Extraction for the U.S. Census Bureau's M3 Survey
Author(s): Osbert Pang*
Companies: U.S. Census Bureau
Keywords:
Abstract:

Signal extraction in economic data frequently entails removing seasonality from the data. This process is typically performed on an individual series basis, with little consideration for the interrelated nature of economic data. A multivariate approach permits more complex structure between series to be accommodated. In this work, we examine the application of multivariate signal extraction techniques to data from the U.S. Census Bureau’s Manufacturers’ Shipments, Inventories, and Orders (M3) survey. We compare the performance to those of univariate model-based signal extraction approaches. We focus on practical implications and discuss findings as they relate to moving from univariate to multivariate signal extraction. This broader overview may provide additional insight into the feasibility of adopting multivariate signal extraction procedures for use in production.


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

Back to the full JSM 2022 program