Multivariate techniques for analyzing and presenting official statistics indicators

Abstract

The production of official statistics is experiencing significant challenges including the handling of massive data sets, the application of computer intensive methods and the integration of data from different sources. Official statistics indicators provide a multivariate perspective, both in form and in content. This perspective requires an implementation of multivariate techniques for data analysis and presentation of findings. The information quality framework is a methodological approach that has been applied to many domain areas including the production of official statistics. Bayesian networks are graphical models that permit decision makers to evaluate alternative scenarios using official statistics. The article presents the information quality framework and discusses a Bayesian network application to Eurostat data. It begins with a background on current official statistics evolutionary changes and concludes with a discussion section that maps some of the challenges of official statistics.