Standardizing Imputation Methods for the Dairy Products Program

• Activity Number: 434
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Darcy Miller*+ and Donnie Fike
• Companies: National Agrigultural Statistics Service and National Agricultural Statistics Service
• Keywords: Imputation ; Missing Data ; Non-Response

Abstract

The National Agricultural Statistics Service's Dairy Products Program estimates the production of butter, cheese, frozen products, dry milk and whey products as well as stocks for major states. Data are collected every month from over 1,000 facilities in the U.S. that produce dairy products. Like most surveys, non-response is an issue and is currently handled using manual imputation, primarily using the previous year's value to impute for the current year. We use computerized imputation methods that mimic the manual process to assess the efficacy of the current process as well as explore alternative computerized machine imputation methods such as a ratio imputation that could account for seasonality as well as overall growth or contraction of production while meeting operational needs. By repeatedly "poking holes" in observed Dairy Products data via several missingness mechanisms, we compare the performance of these methods. Performance is judged by seeing which method most accurately estimated the true dairy production and stocks totals calculated from the complete observed data.