Online Program

Return to main conference page

All Times ET

Program is Subject to Change

Wednesday, June 16
Wed, Jun 16, 10:30 AM - 12:00 PM
TBD
Dealing with Missing and Erroneous Data in Establishment Statistics

Using Multiple Imputation to Mitigate a Sample Reduction (308023)

Darcy Miller, National Agricultural Statistics Service 
*Tara Murphy, National Agricultural Statistics Service 
Benjamin Reist, NASA 

Keywords: editing, imputation, multiple imputation, sample reduction, agriculture,

Like many other agencies and organizations, the U.S. Department of Agriculture’s (USDA) National Agricultural Statistics Service (NASS) needs to produce the same (or better) quality estimates at NASS with less resources. The June Area Survey (JAS) is an annual survey based on an area frame. Data for this survey are collected via in-person interviews and are some of the more labor intensive and expensive interviews to conduct. We assessed the efficacy and feasibility of using imputation in lieu of collecting data from all of the units in the JAS sample while maintaining data quality and providing a measure of reliability associated with imputing some of the data rather than collecting all of it. Benefits realized included cost savings in data collection and analyst editing/imputation. Preliminary case study results applying PROC MI and predictive mean matching imputation methods to a sub-sample of the JAS sample were favorable, so research continued with a simulation study. We present results from this simulation study.