Email), Jacopo Grazzini, (Email), Fabio Ricciato, (Email), Matyas Meszaros, (Email), Konstantinos Giannakouris, (Email), Jean-Marc Museux, (Email), Martina Hahn"> Email)">
14:30 - 15:30
Contributed Paper Session
Room: JENK
Chair:
Per Nymand-Andersen, European Central Bank, Germany, (Email)
Discussant:
Elise Coudin, INSEE, France, (Email)
Organiser:
Jacopo GRAZZINI, European Commission - Eurostat, (Email)
Promoting reproducibility-by-design in statistical offices
Sybille Luhman, (Email), Jacopo Grazzini, (Email), Fabio Ricciato, (Email), Matyas Meszaros, (Email), Konstantinos Giannakouris, (Email), Jean-Marc Museux, (Email), Martina Hahn, (Email)
European Commission - Eurostat, Luxembourg
Because policy advice is becoming increasingly supported by data resources, National Statistical Offices (NSOs) need to leverage the use of new sources of (small or big) data to inform policy decisions. At the time where citizens’ demands for more trust in the public institutions are growing, it underpins the movement towards more open, transparent and auditable (verifiable) decision-making systems. Bearing in mind the reproducibility movement towards Open Science and best practices in the Open Source Software (OSS) community, it is expected that a greater openness, transparency and auditability in designing statistical production processes will result in improved quality of the analysis involved in decision-making as well as increased trust in the NSOs. Drawing upon the Transparency and Openness Promotion guidelines and the Reproducibility Enhancement Principles in computational science, as well as the recommendations to funding agencies for supporting reproducible research and the various calls for open and transparent (data and) algorithms in the field of statistics, we advocate for the following principles: Shared, Transparent, Auditable, Trusted, Participative, Reproducible, and Open (or, in short, STATPRO) to be also adopted by NSOs. These principles build not only on existing and new sources of data, but also on new methodologies and emerging technologies, and advance thanks to innovative initiatives. With increased availability of open data, new developments in open technologies and open algorithms, as well as recent breakthroughs in data science, it is believed that they can help improve current governance processes by enabling data-informed evidence-based decision-making and potentially reduce the bias, costs and risks of policy decisions. In this regard, Official Statistics should be accompanied by access to the data analysed whenever possible, the detailed metadata information, the underlying assumptions (models and methods) and also the tools (software) used to generate them. In the context of a “post-truth” society, the STATPRO principles present substantial promises for Citizen Statistics and e-Official Statistics, e.g. for the interaction between NSOs, data users and data producers. Indeed, they make it possible to provide the public with the ability to both perform the analysis and repeat it with different hypotheses, parameters, or data, hence translating policy questions into a series of well-understood computational methods and scrutinizing the final decision. In this contribution, we further emphasize the need for Official Statistics to go beyond current practice and exceed the limits of the NSOs and the European Statistical System so as to reach and engage with produsers – e.g. statisticians, scientists and also citizens. Through the adoption of some best practices derived from the OSS community and the integration of modern technological solutions, the STATPRO principles can help create new participatory models of knowledge and information production. We illustrate this trend through Eurostat recent initiatives.


Reference:
CPS10-002
Session:
Open, transparent and reproducible
Presenter/s:
Sybille Luhman
Presentation type:
Oral presentation
Room:
JENK
Chair:
Per Nymand-Andersen, European Central Bank, Germany, (Email)
Date:
Thursday, 14 March
Time:
14:30 - 15:30
Session times:
14:30 - 15:30