Multiple Imputation for Parametric Inference Under a Differentially Private Laplace Mechanism

Thumbnail Image

Files

RRS2019-05.pdf (483.18 KB)

Links to Files

https://www.census.gov/srd/papers/pdf/RRS2019-05.pdf

Permanent Link

http://hdl.handle.net/11603/14969

Collections

UMBC Mathematics and Statistics Department
UMBC Faculty Collection

Author/Creator

Author/Creator ORCID

Date

2019-05-09

Type of Work

reports
Text

Department

Program

Citation of Original Publication

Martin Klein and Bimal Sinha, Multiple Imputation for Parametric Inference Under a Differentially Private Laplace Mechanism, Research Report Series, 2019, https://www.census.gov/srd/papers/pdf/RRS2019-05.pdf

Rights

This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
This is a work of the United States Government. In accordance with 17 U.S.C. 105, no copyright protection is available for such works under U.S. Law." in either case, put on a public domain creative commons license.
Public Domain Mark 1.0

Subjects

Differential Privacy
EM Algorithm
Multiple Imputation
Parametric Model
Statistical Disclosure Control

Abstract

In this paper we consider the scenario where continuous microdata have been noise infused using a differentially private Laplace mechanism for the purpose of statistical disclosure control. We assume the original data are independent and identically distributed, having distribution within a parametric family of continuous distributions. We employ a modification of the standard Laplace mechanism that allows the range of the original data to be unbounded. We propose methodology to analyze the noise infused data using multiple imputation. This approach allows the data user to analyze the released data as if it were original, i.e., not noise infused, and then to obtain inference that accounts for the noise infusion mechanism using standard multiple imputation combining formulas. Methodology is presented for univariate data, and some simulation studies are presented to evaluate the performance of the proposed method. An extension of the proposed methodology to multivariate data is also presented.