New Challenging Problems in SAE from Real Life


Chair: Daniel Bonnery

This session will focus on practical issues encountered in small area estimation that are new in the sense that they have been tackled in some recent work only. Three original topics will be addressed, to show the diversity and importance of the issues that still require theoretical and methodological research in small area estimation. The topics are related to prediction of functional data for small area, robustness of estimation method and accounting for measurement error, and have been brought up by practical studies on real data. The different statisticians from University, Government agency and Private companies who will intervene in this session will provide high level theoretical presentation and will pay particular attention to presenting the real and important problem they had to solve. First proposed presentation will be given by Anne de Molinier, who works in EDF, the French Electricity Company, and has to predict functional data (in this example, electricity consumption function of time on small areas), model choice and prediction is chalenging for such data. Second presentation will be given by Jan van den Brakel (Statistics Netherlands and Maastricht Unversity School of Business and Economics) who will discuss the use of a multivariate structural time series modelling approach to improve the precision of direct estimates with sample information from previous periods and auxiliary series derived from related (big) data sources. Third presentation will be given by Laura Dumitriescu (University at Wellington, New Zealand), who proposed a method to account for the measurement or sampling error on auxiliary information used for prediction in a small area estimation context.


Social media as a data source for official statistics; the Dutch Consumer Confidence Index

One way to use big data sources in the production of official statistics is to use them as auxiliary information in models for small area estimation procedures. Marchetti et al. (2015) used mobility data to predict poverty in a Fay Herriot model that improves the effective sample size with sample information from other domains. Most […]

This session was organised by Valentin Patilea and Daniel Bonnery.