Constructing Strata of PSUs for the Residential Energy Consumption Survey

• Activity Number: 606
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Rachel Harter* and Patrick Chen and Joseph McMichael and Edgardo Cureg and Samson Adeshiyan and Katherine Morton
• Companies: RTI International and RTI International and RTI International and Energy Information Administration and Department of Energy and RTI International
• Keywords: strata ; decision trees ; principal components ; cluster analysis

Abstract

Textbooks provide guidance on the general principles and desirable properties of defining sampling strata. This paper reviews some basic exploratory methods for determining stratification variables, including principal component analysis, cluster analysis, correlations, regression analysis, and decision trees for reducing the set of potential variables. Although all stratification methods use auxiliary variables available for the entire frame, the decision tree, regression, and correlation approaches also use prior outcome data, which may be available for just a sample of units. The principal component method combined with cluster analysis, on the other hand, focuses on relationships among stratification variables. Using both principal components/cluster analysis and decision trees, we stratify primary sampling units for the Residential Energy Consumption Survey and compare the resulting strata.