Affiliations: Department of Statistical Science, Duke University, Durham, NC, USA
Correspondence:
[*]
Corresponding author: Jerome P. Reiter, Department of Statistical Science, Duke University, Box 90251, Durham, NC 27708, USA. Tel.: +1 919 668 5227; Fax: +1 919 684 8594; E-mail:jerry@stat.duke.edu
Abstract: We present approaches to generating synthetic microdata for
multivariate data that take on non-negative integer values, such as magnitude data in
economic surveys. The basic idea is to estimate a mixture of Poisson
distributions to describe the multivariate distribution, and release
draws from the posterior predictive distribution of the model. We
develop approaches that guarantee the synthetic data sum to marginal
totals computed from the original data, as well approaches that do not
enforce this equality. For both cases, we present methods for assessing disclosure
risks inherent in releasing synthetic magnitude microdata. We
illustrate the methodology using economic data from a survey of
manufacturing establishments.
