Read more
This Open access book gives an overview of current research and developments on the incorporation of machine learning in official statistics. It covers methodological questions, practical aspects and cross-cutting issues.
Machine learning has become an integral part of official statistics over the last decade. This is evident in its many applications in numerous countries and organisations. At the same time, the integration of machine learning into statistical production raises questions about the right mathematical and statistical methodology, the consideration of quality standards and the appropriate IT support. In its four sections, "Methodological aspects", "Legal, ethical, and quality aspects", "Technological aspects" and "Use cases and insights", the book highlights current developments, provides inspiration, outlines challenges and offers possible solutions. It is aimed at methodologists in statistical offices and comparable institutions as well as scientists who are concerned with the further development and responsible use of machine learning
List of contents
Introduction.- 1. ML in official statistics (T Augustin, AL Boulesteix - LMU Munich).- 2. Evaluation of generalization error (B Bischl, AL Boulesteix, R Hornung, H Kümpel, S Fischer, A Bender, L Bothman, L Schneider -- LMU Munich).- 3. ML and Design of Experiments/Sample size calculation (T Augustin - LMU Munich).- 4. Interpretable ML (B Bischl, L Bothmann, S Dandl, G Casalicchio -- LMU Munich).- 5. Set-valued methods for ML in official statistics (T Augustin - LMU Munich).- 6. Ethics and Fairness (F Kreuter - at LMU Munich).- 7. Quality aspects of ML (Y Saidani et al -- Statistical Offices in Germany).- 8. A statistical matching pipeline (T Küntzler --- Destatis).- 9. Legal Aspects of ML (T Fetzer - Mannheim University).
About the author
Florian Dumpert heads a division at the Federal Statistical Office of Germany that develops methodological and technological solutions and architectures for statistics production. The focus of his work is on the quality-assured integration and use of machine learning for the purpose of digitalisation, standardisation and automation of official statistics. His research interests include statistical machine learning, statistical data processing and imputation. He regularly participates in national and international projects on these topics and represents the disciplines in relevant working groups and committees.
Summary
This Open access book gives an overview of current research and developments on the incorporation of machine learning in official statistics. It covers methodological questions, practical aspects and cross-cutting issues.
Machine learning has become an integral part of official statistics over the last decade. This is evident in its many applications in numerous countries and organisations. At the same time, the integration of machine learning into statistical production raises questions about the right mathematical and statistical methodology, the consideration of quality standards and the appropriate IT support. In its four sections, "Methodological aspects", "Legal, ethical, and quality aspects", "Technological aspects" and "Use cases and insights", the book highlights current developments, provides inspiration, outlines challenges and offers possible solutions. It is aimed at methodologists in statistical offices and comparable institutions as well as scientists who are concerned with the further development and responsible use of machine learning
