Meeting notes - September 2022

[unece-stat]

Participants: Jenny Linnerud, Zoltan Vereczkei, Juan Munoz, Cecilia Casagrande, Andrea Petres, Florian Vucko, Ayman Hathoot, Christophe Jones and Wai Kit (Jackie) Si Tou

Issue #17 Registers in statistical production: their roles and types, and how to model them in GSIM
- Action: Zoltan to set up a separate meeting with others and continue discussion. Issue is whether to add qualifier “statistical” and how much we want to differentiate “statistical register” and “administrative register”.

Issue #6 Missing GSIM class – quality indicators

• On the new version: it has simplified the current version and covers the examples we have. Regarding Referential Metadata Subject – to keep it for now?
• Are there reasons why Referential Metadata Set should be modelled differently than Data Set? If there is no fundamental difference, can we not just use controlled vocabulary to indicate whether information refers to data or metadata?
• Under Issue Missing GSIM class – specification of Exchange Channel #5, it was proposed to introduce a new class “Information Structure” which can be superclass of Data Structure and Metadata Structure. If there are any commonalities in the way we structure data and metadata, they can be moved to this superclass.
• Currently, referential metadata has Referential Metadata Attribute and Referential Metadata Subject, but it might be just because we don’t think structure information as variables (as it is in data). If we start doing, then we can represent Referential Metadata Attribute as variable and describe data and metadata in the same way.
• Action: Flavio to create issue on the alignment of the ways metadata and data are structured in GSIM
• Whether something is metadata or data depends on the context. Data of someone can be metadata of someone else. Also, depending on the format, metadata can be data (e.g., if we change short format data to long format, the variable names become data points). Having separate and different classes for metadata and data would make such change more difficult. We need to look at GSIM structure area with this perspective, it may not be feasible to introduce big changes during this revision, but finding commonalities and putting them in the superclass would be a starting point
• One way to differentiate metadata and data is that the former is normally not confidential, while with the later, you might need to be more careful (exception includes when metadata contains information on confidentiality algorithm).

Issue #23 Update UML EA model

• Enterprise Architect (EA) file had been produced for each new version of GSIM.
• Having EA is useful as we have one interconnected model, helpful in maintaining the cardinality and association, deriving diagrams, etc.
• Changes so far have been made in silos, EA will allow us to check the consistency, etc.

Issue #16 Missing workflow class(es) – Concrete/detailed representation of BusinessProcess: may not have time to go over during this revision

Issue #14 GSIM Feedback from Hungarian Central Statistical Office: to go over this issue in the next meeting

Issue #12 Information management datasets to check if new referential metadata model would work on this example

Issue #10 GSIM Implementation Guide

• The use cases that we have covered in different issues would be useful for the users. Need to compile them
• Providing the context of statistical process (GSBPM) is helpful
• Rather than definitions and explanatory texts, providing explanation to help understanding how to use some classes (e.g., referential metadata)