Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number:
Register
172 - Government Statistics Section P.M. Roundtable Discussion (Added Fee)
Type: Roundtables
Date/Time: Tuesday, August 10, 2021 : 12:00 PM to 1:20 PM
Sponsor: Government Statistics Section
Abstract #317504
Title: Toward Developing Viable Missing Data Treatments for Short-Term Statistics During a Pandemic
Author(s): Stephen Kaputa*
Companies: U.S. Census Bureau
Keywords: survey nonresponse; time series; missing data treatment; imputation
Abstract:

The production of timely and accurate short-term statistics is a key function of many National Statistical Offices. These indicators are used by policy makers, professional organizations, and financial markets to monitor an economy in real time. Consequently, the collection cycle is usually monthly or quarterly, leaving little time for nonresponse follow-up as the measures being output soon after collection. In general, short-term statistics comprise time series, and missing data treatments often rely on assumptions of autoregression and seasonality. Beginning in March 2020, many short-term statistics programs had unprecedented levels of nonresponse, with apparent breaks in long-time stationary series. In other words, the formerly reliable missing data treatment procedures were no longer viable. In this session, we will discuss general issues of missing data treatment during an economic upheaval, then focus on specific issues such as imputation of zero values versus positive values. Other discussion points will include imputation methods used in practice, both successfully and unsuccessfully.


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

Back to the full JSM 2021 program