Why Shuffle When You Can Use Robust Statistics for SDC - A Simulation Study

Abstract. The aim of this study was to compare different microdata protection methods for numerical variables under various conditions. Most of the 21 methods used in this paper have been implemented in the R-package sdcMicro which is available for free on the comprehensive R archive network (http://cran.r-project.org). The rest of the methods used can easily applied within other R-packages. While most methods work well for homogeneous data sets, some methods fail completely when the confidential variables contain outliers which is almost always the case with data from official statistics. To overcome these problems classical methods such as microaggregation or methods based on regression like shuffling have been robustified. The methods were tested on different bivariate data sets with contaminations. In addition to that, a simulation study was performed to test the methods under different outlier scenarios.