English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

One of several explanations of why Homo Sapiens is the only surviving sub-species of the genus Homo is the extended length of our childhood and adolescence. The value of this extended maturation and developing period may be that it allows us to learn and carry out complex tasks. Like Homo Sapiens, methodology for synthetic data has had a long learning period. The idea of using synthetic data for disclosure control was conceived more than 20 years ago [1–3], but it was a further 10 years before the first papers describing how to do it appeared in the literature [4,5]. The subsequent decade was one of rapid development and innovation when the methodology was tested and expanded. The energy and enthusiasm for synthetic data of Reiter and his colleagues was responsible for many major developments; see the monograph by Drechsler [6] for a review. Towards the end of synthetic data’s second decade real applications began to appear [7–9]. Two of the four substantial papers that deal with synthetic data in this issue [10,11] are examples of mature methodology, while the other two [12,13] deal with disclosure control, the aspect of synthetic data that is at an early stage in its development. My comments here are from the point of view of a practitioner looking for useful and workable ideas in this field. Our project to provide data for the UK Longitudinal Studies (LSs) is referred to in Vilhuber et. al.’s overview of international developments [14]. More details of our methods and our synthpop package for R are available [15–17]. The paper by Vilhuber and Miranda [10] is a particularly good example of a new method developed in response to a real need. It is rare for a journal article to illustrate the difficulties that arise when statistical procedures are used on real data with all their imperfections and idiosyncrasies. The idea in this paper is to produce tabulations from synthetic data and use them to replace cells in tables from the original data that have been suppressed because of perceived disclosure risk. Suppressed cells in the open-source tabulations, Business Dynamics Statistics1 (BDS), are replaced with those from tabulations of the synthetic Longitudinal Business Data (synLBD).2 The authors take advantage of the fact that the BDS and the synLBD are both derived from the same original data sources. Their tentative conclusions are somewhat disappointing, showing that using the synLBD does not improve validity compared to a much simpler method of adding multiplicative noise to the cells. But the reasons for this relate to known problems with the utility of the synLBD for the very statistics that the BDS tabulates, and also to differences between the versions of the BDS and synLBD that were available to the authors. The problems with the synLBD have been remedied by a new methodology [18] that will be used for future versions. Their methods may work better when the synthetic data are produced specifically for the purpose of replacing cells in tables. In particular, the methods based on their algorithm 2, which overcome problems of inconsistent

Comments on four papers on synthetic data in Volume 32 Issue 1 the Statistical Journal of the IAOS