Search

The twin GRACE-A and -B satellites are in orbit since more than 6 years and still perform well. They deliver gravity data of unprecedented accuracy, which enables hydrologists, oceanographers and geophysicists to study the temporal variations of the Earth΄s gravity field. But the baseline accuracy computed in a pre-launch simulation study (Kim, 2000) has not yet been reached by a factor of about 15. We therefore have to raise the question: are there improvements in the processing of GRACE data possible? To answer this question, a simulation study was performed, using the same software, processing strategy, background models and standards used at GFZ for the analysis of real GRACE data. We present the results and analysis of this simulation study. Initially a closed loop simulation shows, that GFZ΄s EPOS software is numerically stable. The GRACE orbit geometry is sufficient and the sampling adequate to solve for monthly gravity fields at least up to degree and order 150. The estimation of instrument parameters as suggested by Kim (2000) does not absorb the gravity signal, but greatly reduces systematic effects in the observations. The accelerometer noise proved to be an important reason for not reaching the baseline accuracy with the processing strategy used so far. Additional accelerometer parameters do not really help, but the shortening of arcs gives promising results. Different background model errors were introduced and the ocean tide model turned out to be a probable error source, while atmospheric tides play a minor role. [ABSTRACT FROM AUTHOR]

The aim of this study is to evaluate the risk of re-identification related to distance-based disclosure risk measures for numerical variables. First, we overview different - already proposed - disclosure risk measures. Unfortunately, all these measures do not account for outliers. We assume that outliers must be protected more than observations near the center of the data cloud. Therefore, we propose a weighting scheme for each observation based on the concept of robust Mahalanobis distances. We also consider the peculiarities of different protection methods and adapt our measures to be able to give realistic measures for each method. In order to test our proposed distance based disclosure risk measures we run a simulation study with different amounts of data contamination. The results of the simulation study shows the usefulness of the proposed measures and gives deeper insights into how the risk of quantitative data can be measured successfully. All the methods proposed and all the protection methods plus measures used in this paper are implemented in R-package sdcMicro which is freely available on the comprehensive R archive network (http://cran.r-project.org). [ABSTRACT FROM AUTHOR]

In the release of microdata files, reidentification of a record implies disclosure of the values of a possibly large set of sensitive variables. When microdata files are released by statistical Agencies, a careful assessment of the associated disclosure risk is therefore required. In order for an informed decision to be made, maximising accuracy and precision of the risk estimators is crucial. Clearly such characteristics will affect the risk assessment process and Agencies should choose the estimator that performs best. In fact, estimators may perform poorly, especially for those records whose real risk is higher. To improve estimation, we propose to introduce external information, arising from a previous census as is done in the context of small area estimation (see [10]). In [4] we considered SPREE - type estimators that use the association structure observed at a previous census (see [9]); in this paper we consider models that use the structure of a population contingency table while allowing for smooth variation of the latter. To assess the statistical properties of this estimator and compare it with alternative approaches, we show results of a simulation study that is based on a complex sampling scheme, typical of most households surveys in Italy. Comparison is made with a simple SPREE estimator and a Skinner-type estimator [13,6], applied to a complex sampling scheme. [ABSTRACT FROM AUTHOR]

The release of survey microdata files requires a preliminary assessment of the disclosure risk of the data. Record-level risk measures can be useful for ˵local″ protection (e.g. partially synthetic data [21], or local suppression [25]), and are also used in [22] and [16] to produce global risk measures [13] useful to assess data release. Whereas different proposals to estimating such risk measures are available in the literature, so far only a few attempts have been targeted to the evaluation of the statistical properties of these estimators. In this paper we pursue a simulation study that aims to evaluate the statistical properties of risk estimators. Besides presenting results about the Benedetti-Franconi individual risk estimator (see [11]), we also propose a strategy to produce improved risk estimates, and assess the latter by simulation. The problem of estimating per record reidentification risk enjoys many similarities with that of small area estimation (see [19]): we propose to introduce external information, arising from a previous census, in risk estimation. To achieve this we consider a simple strategy, namely Structure Preserving Estimation (SPREE) of Purcell and Kish [18], and show by simulation that this procedure provides better estimates of the individual risk of reidentification disclosure, especially for records whose risk is high. [ABSTRACT FROM AUTHOR]