Panel: Metadata for Statistical Data

Let us consider the number, 3.6. The lack of context around it – what does it mean, how was it produced, what is its accuracy, etc. – renders it unusable. The answers to these questions (and others) provide the information needed to understand what the number represents. Giving a hint, this number was recently released by the US Bureau of Labor Statistics; this and the answer to the questions are statistical metadata. They are metadata because they describe the number, designs, and processing leading to its creation, and they are statistical because the number was produced by a process that uses probability and statistics. This talk will describe the activities that lead to the production of the number, providing an introduction to many aspects of statistical metadata. Along the way we will mention the many standards that address the metadata needs for statistical data production.

First released in 2002, the Statistical Data and Metadata eXchange standard (SDMX) started as “an initiative to foster standards. The standard has been further enhanced to include other statistical data and metadata use cases, for example making authoring, validation and exchange easier, faster, and with better quality. SDMX is not only a set of technical standards but also includes content-oriented guidelines that aim at harmonizing metadata across statistical domains and agencies. An example is a guideline on how to perform structural modelling of datasets.

This presentation will provide an overview on SDMX. Also, it will describe how the OECD is using SDMX to harmonise their structural metadata, metadata governance, and tools collaboration at the OECD and with other agencies.

Metadata plays a key role in statistical production: from concepts, classifications and variables to retention and provenance information, metadata is created, used and shared across all phases of the data lifecycle. As we move towards a FAIRer infrastructure (Findable, Accessible, Interoperable, Reusable), metadata management needs to leave behind its silo-centric, tool-specific history and embrace a new, integrated and cross-domain approach. Statistics Canada is in the early stages of implementing a virtual metadata integration platform, the Metadata Hub, at the core of which stands a semantic infrastructure anchored in an RDF triple store. This triple store integrates content from a number of metadata repositories, e.g., Colectica, Ariā, SDMX .Stat Toolkit, GeoNetwork and CKAN across a number of standards, e.g. DCAT, DDI, SDMX, and S/XKOS, among others. We’ll discuss here our experience so far, use cases and the way forward.