Bayesian marked point process modeling for generating fully synthetic public use data with point-referenced geography

Abstract

Many data stewards collect confidential data that include fine geography. When sharing these data with others, data stewards strive to disseminate data that are informative for a wide range of spatial and non-spatial analyses while simultaneously protecting the confidentiality of data subjects' identities and attributes. Typically, data stewards meet this challenge by coarsening the resolution of the released geography and, as needed, perturbing the confidential attributes. When done with high intensity, these redaction strategies can result in released data with poor analytic quality. We propose an alternative dissemination approach based on fully synthetic data. We generate data using marked point process models that can maintain both the statistical properties and the spatial dependence structure of the confidential data. We illustrate the approach using data consisting of mortality records from Durham, North Carolina.

Generating Partially Synthetic Geocoded Public Use Data with Decreased Disclosure Risk by Using Differential Smoothing

“…One approach to do so would be that of Quick et al . (), which uses log‐Gaussian Cox processes (LGCPs) (Møller et al ., ) with an underlying spatial structure to model home addresses within a marked point process model. By doing so, the synthesizer of Quick et al .…”

“…For the data in distribution (1), this would not only require generating synthetic n × 1 vectors Y †.l/ , but it would also require constructing a model from which to generate a collection of synthetic locations S †.l/ and any other individual level attributes. As described in Wang and Reiter (2012) and Quick et al (2015), however, approaches for generating fully synthetic point-referenced data sets can be quite computationally burdensome. Thus, in some instances, it may be attractive to take a partially synthetic approach in which only a collection of values or variables are replaced with imputed values (e.g.…”

“…As described in Wang and Reiter () and Quick et al . (), however, approaches for generating fully synthetic point‐referenced data sets can be quite computationally burdensome. Thus, in some instances, it may be attractive to take a partially synthetic approach in which only a collection of values or variables are replaced with imputed values (e.g.…”

Zeros and ones: a case for suppressing zeros in sensitive count data with an application to stroke mortality

“…In order to evaluate both the disclosure risk and the utility of the suppressed data, we consider the use of synthetic data (e.g. Little, ; Kennickell, ; Reiter, ; Quick et al, ). Specifically, we let θ = ( β 0 , Z T , ϕ T , σ 2 , τ 2 ) T and generate synthetic values, , using the posterior predictive distribution, based on the hierarchical model in .…”

Establishing a framework for Open Geographic Information science

“…For example, a variance-covariance matrix might be used to generate new data that serves as a proxy for the original data. Research into 'spatial data synthesizers', such as Quick et al (2015), would thus be enormously beneficial. The US Census Bureau has started publishing synthetic individual level data based on highly sensitive administrative data from agencies like the IRS and the Social Security Administration (Bureau 2014).…”