Releasing Synthetic Microdata for Magnitude Tabular Data

• Activity Number: 355
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 11:35 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Lan Wei*+ and Jerome P. Reiter
• Companies: Duke University and Duke University
• Keywords: Data Confidentiality ; Mixture Model ; Nonparametric Bayesian

Abstract

Establishment microdata is subjected to strict confidential protection. Releasing it while providing meaningful statistical inference is a challenge. This is especially true for micro discrete/count data with fixed marginal totals. We develop a new statistical disclosure limitation (SDL) technique based on data synthesis. We propose a mixture of Poisson distributions for multivariate count data and a synthesis method subject to fixed marginal totals. We apply the procedure to a data from the Colombian Annual Manufacturing Survey. The results show that the synthetic data can provide high utility. We also present a method for disclosure risk assessment.