Statistical Model of the 2001 Czech Census for Interactive Presentation
Jiří Grim, Jan Hora, Pavel Boček, Petr Somol, Pavel Pudil
This article describes the application of a recently developed method of interactive statistical database presentation to the 2001 Czech Census. The method is based on estimating the multivariate probability distribution of the original microdata, which are supposed to be discrete or discretized continuous. The estimated statistical model in the form of a distribution mixture of product components can be used as a knowledge base of a probabilistic expert system. By means of the probabilistic inference mechanism we can derive conditional distributions of variables for each subpopulation interactively without any further access to the source database. The conditional distributions (histograms) describing the properties of subpopulations represent the basic form of user information. The statistical model does not contain the original data and therefore can be distributed without any confidentiality concerns. The accuracy achievable by the statistical model is comparable with that of the anonymized subsets of microdata.
Interactive statistical model, census data presentation, distribution mixtures, data modeling, EM algorithm, incomplete data, data reproduction accuracy, data mining