Evaluating the Impact of Imputation on RECs Data

• Activity Number: 176 - Contributed Poster Presentations: Section on Statistics and the Environment
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Sarrynna Sou* and Peter Frechtel and Amang Sukasih and Katie Lewis and James Berry and Victoria Scott
• Companies: RTI International and RTI International and RTI International and U.S. Energy Information Administration and U.S. Energy Information Administration and RTI International
• Keywords: item nonresponse ; missing data ; energy ; imputation ; evaluation ; RECs

Abstract

The 2015 Residential Energy Consumption Survey (RECS) was conducted using CAPI, Web, and Mail/Paper surveys. Both the RECS CAPI and Web/mail surveys have missing data due to item nonresponse. Because item nonresponse can cause bias, a process that we have undertaken is the Cyclical Tree-Based Hot Deck (CTBHD) imputation method (Wine, Bryan, & Siegel, 2014). This poster will explore the following two questions: Did this imputation method make a difference? How successful was it? The visual will provide insight on how effective the imputation processes were on the data. Through a series of estimation comparisons, statistical tests, and graphics of various variables, the poster will compare the non-imputed dataset with the imputed dataset, as well as with the external dataset. After reviewing the impact, the poster will explore the quality of the imputation. In order to assess the quality, the poster will set forth questions on what is expected of the imputation and what it has yielded. The poster's objective is to further the discussion on how analysts should establish a technique to assess imputation through evaluating the RECS imputation impact and literature review.