Using Statistical Models for Sample Design of a Reinterview Program

• Activity Number: 365
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Jianzhu Li*+ and Michael J. Brick and Phyllis Singer and Bac Tran
• Companies: Westat, Inc. and Westat, Inc. and U.S. Census Bureau and U.S. Census Bureau
• Keywords: quality control ; falsification ; curb stone ; rare event

Abstract

The Census Bureau regularly conducts a large number of major demographic surveys. It has an ongoing Quality Control (QC) program which relies on a reinterview program as the primary method to evaluate field work or monitor the work of the field representatives (FRs), with a special focus on identifying falsified and misclassified interviews. Currently only about 2% of the original interviews in the Current Population Survey (CPS) are assigned for QC reinterview, and the sampling of QC reinterviews does not rely on any information other than interviewer's experience level. To improve the ability to detect falsified cases, we fit a logistic regression model to predict the likelihood of falsification using the data from the original interviews as the predictor variables. The fitted model was then used to construct two alternative sampling schemes for the QC reinterview.