ENBES webinar
Audit Sampling as a New Quality Standard for Statistics Production
Tuesday, November 12, 2024, 15:00 - 17:30 CET
- - hosted virtually by Statistics Portugal -
Producing statistics from transactions data, administrative data, and other sorts of "found" data - as opposed to "designed" data purposefully collected in surveys - is becoming a new reality for many producers of statistics. However, this contrasts with foundations on which the scientific basis for official statistics has been built in the 20th century, ultimately rooted in the sampling design that is both known (hence, transparent) and reproducible. Statistics producers that use found data still face the need and obligation to provide reliable information to users and their governing bodies on the quality (in particular, accuracy) of the produced statistics. Quality frameworks and guidelines have been produced for multisource statistics regarding the various error sources and statistical processes. Yet a consensus or directive is lacking whether one can accept for official statistics any models (or methods) and, if so, how one can validate such a model (or method) in a transparent and reproducible manner (i.e. as in the case of survey sampling).
To highlight this increasingly relevant issue, and to indicate a way forward, ENBES is organising a webinar entitled "Audit sampling as a new quality standard for statistics production", to be held on November 12, 2024. To put it short, by auditing sampling, one would apply sampling for the purpose of assessing the errors of any given statistics, instead of producing such statistics by survey sampling in the first place, such that the associated uncertainty measures are valid with respect to the audit sampling design, regardless of the veracity of the models or methods (whichever they are) that have been used to produce the given statistics. The webinar will feature two speakers, Prof. Li-Chun Zhang (Statistics Norway and University of Southampton) and Dr Sander Scholtus (Statistics Netherlands).
Prof. Zhang will introduce an audit sampling framework for multisource statistics and some associated uncertainty concepts that may be particularly relevant for big-data estimators. Real-life examples of audit sampling will be used to highlight their intuition and public acceptance, whether or not these undertakings were historically conceived or presented as audit sampling. Dr Scholtus will present a case study in which Statistics Netherlands has applied audit sampling to evaluate quality of administrative data and input the evaluations into the production process in order to improve the quality of produced statistics. This case study focuses in particular on evaluating the effect of classification errors on produced statistics. The presentations will be followed by Q & A and discussion.
Registration closed. To register, please follow this link: https://forms.gle/vNiW6BXksNQeEsJf6.
Teams link to participate in the webinar: will be provided to registered participants.
To contact the webinar organiser, please write to: info@enbes.org.