Generalized Linear Models with Variables Subject to Post Randomization Method

• Activity Number: 665
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Yong Ming Jeffrey Woo*+ and Aleksandra Slavkovic
• Companies: Penn State University and Penn State University
• Keywords: generalized linear model ; estimation ; perturbation ; statistical disclosure control ; EM algorithm ; utility

Abstract

The main goal of Statistical Disclosure Control (SDC) methodology is to provide society with access to confidential data such that individual information is sufficiently protected against disclosure and at the same time data utility is preserved for valid statistical inference. The Post Randomization Method (PRAM) advocates release of perturbed data, where values of categorical variables are perturbed via some known probability mechanism. Estimation with perturbed data without accounting for PRAM leads to biased estimates, hence raising issues with data utility. To address these issues, we propose a number of EM-type algorithms to obtain unbiased estimates of generalized linear models (GLMs) fitted to perturbed data. A few measures of disclosure risk will also be evaluated and discussed, as well as applications of the proposed methodology to data from the 1993 Current Population Survey.