Initial Research on Computing Estimated Measures of Sampling Variability for the Weekly Natural Gas Storage Report

• Activity Number: 620
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): David Kinyon* and Samson Adeshiyan and Joseph Conklin and Jose Villar
• Companies: Energy Information Administration and Energy Information Administration and Energy Information Administration and Energy Information Administration
• Keywords: establishment surveys ; complex surveys ; bootstrap ; variance estimation ; nonlinear estimator

Abstract

The U.S. Energy Information Administration releases the Weekly Natural Gas Storage Report, designated as a Principal Federal Economic Indicator, to provide weekly estimates of both working gas levels held in underground storage facilities and the net changes in these levels, using survey data collected from a sample of natural gas storage operators in the Lower 48 states. Because the estimator used to compute these estimates is nonlinear, initial research was conducted to compute estimated measures of sampling variability using the bootstrap method. The research focused on the weekly estimates from April to November of 2015 that were based on the most recent sample that was selected and the new five-region breakout. Based on this initial research, it appeared that accurate measures of sampling variability could be estimated using a repeatable process, which could be implemented in production to enhance the content of future releases. This paper will discuss the nonlinear estimator, implementation of the bootstrap method, and the results of the research to date on computing estimated measures of sampling variability for both level and change estimates produced from the survey.