Johann Bacher

A Probabilistic Clustering Model for Variables of Mixed Type

This paper develops a probabilistic clustering model for mixeddata. The model allows analysis of variables of mixed type: thevariables may be nominal, ordinal and/or quantitative. The modelcontains the well-known models of latent class analysis as submodels.As in latent class analysis, local independence of the variables isassumed. The parameters of the model are estimated by the EMalgorithm. Test statistics and goodness-of-fit measures are proposedfor model selection. Two artificial data sets show the usefulness ofthese tests. An empirical example completes the presentation.

Re-identifying register data by survey data using cluster analysis: an empirical study

More and more empirical researchers from universities or research centres like to use register or survey data collected by statistical agencies or the social security system, since these data can by used for several empirical studies, e.g. the analysis of special groups or quantitative effects of economic or social policies. Most of the data required have to be (factually) anonymised before they are disseminated to preserve confidentiality. In the area of statistics on households and individuals this path has been pursued in Germany for several years. The transmission of de facto anonymised datafiles has proved to be a good form of co-operation between scientists and statisticians.Factual anonymity of the data depends on the costs and benefits of a potential reidentification. The paper assumes that the intruder only accepts low costs. Therefore he uses a cluster analysis module that is available in a standard statistical software package to re-identify persons. After a description of the method different factors influencing the re-identification risk are studied using German employment statistics (register data) and the German Life History Study (survey data). The factors are: sample fraction and number of (irrelevant) variables. The results show, that the number of identifiable persons is remarkable high. Furthermore it can be confirmed with the cluster analysis that the number of re-identifiable records increases with increasing sampling fraction and that irrelevant variables reduce this number.

Sozialberichterstattung über die Armutsgefährdung von Kindern

Die Forderung nach einer Sozialberichterstattung uber die Armutsgefahrdung von Kindern beinhaltet einen wichtigen Trend, der sich innerhalb der Sozialberichterstattung bereits seit den 90er-Jahren abzeichnet: die Spezialisierung der Berichterstattung auf ausgewahlte Zielgruppen und/oder bestimmte gesellschaftspolitische Bereiche. So existieren Berichtssysteme uber spezifische Lebens- und Politikbereiche (vgl. Noll 1997) wie Familie, Gesundheit, Armut oder Umwelt und uber bestimmte Bevolkerungsgruppen (Senioren, Familien oder Frauen), sie weisen jedoch einen unterschiedlichen Entwicklungsstand auf.