Statistical Engineering Division SeminarSynthetic Microdata Simulation for Confidentiality Protection Using Regression Quantiles
Jennifer Huckett Abstract Government agencies must simultaneously maintain confidentiality of individual records and disseminate useful microdata. Iowa's Legislative Services Agency (LSA) needs predicted state tax revenue based on proposed policy changes calculated from individual income tax returns. Iowa's Department of Revenue (IDR) cannot provide individual records to LSA by law. Currently, LSA submits requests to IDR that IDR computes and reports to LSA. This is inefficient for both agencies. We study options for IDR creating a synthetic tax return file for release to LSA. In this talk I will discuss combining quantile regression, hot deck imputation, and additional confidentiality-preserving methods to produce releasable, usable data. Measures of disclosure risk are considered. In addition to other measures of data utility, several versions of microdata can be multiply imputed to assess uncertainty. NIST Contact: John Lu, (301) 975-3208.
Date created: 8/15/2007 |