A Radical Proposal for the Provision
of Micro-Data Samples and
the Preservation of Confidentiality
Stephen E. Fienberg
The problem of the preservation of confidentiality for the release of
micro-data samples is reconsidered in the context of estimating an
empirical cumulative density function. A proposal is made for the use
of a bootstrap-like approach to the generation of synthetic microdata
files. Many details in the proposal require careful attention, and the
implications of the proposed method for data disclosure still need to
be explored empirically. The method suggested here bears a remarkable
similarity to a proposal by Rubin (1993) for the use of multiple
imputation for data disclosure limitation.
Keywords: Bootstrap; Data-disclosure limitation; Multiple imputation;
Simulated micro-data sets; Smoothed cumulative distribution function.
Here is the full postscript text for this
technical report. It is 77 kbytes.