Value Doamain has the following definition:

"The permitted range of values for a characteristic of a variable"   Why "characteristic of a variable", isn't Value Domain "the permitted range of values for a variable"? Or is the definition supposed to be" The permitted range of values for a characterictic of a population"?  Characteristic is not defined within GSIM, it would be useful if it was.

 

  • No labels

2 Comments

  1. user-8e470

    Meeting 10/1:  "Characteristic of a variable" does not make sense. It should be "the set of permissible values for a variable"

     

    ACTION: Change definition

  2. Alistair Hamilton

    I agree that is the simplest change.

    Underlying the existing wording, however, might be some residual thinking from ISO 11179 Part 3 that we've missed being present in GSIM when it comes to implementation for GSIM at the ABS.

    As I think I shared during the CSPA LIM Variables sprint, the following starts with a comparison between an ISO 11179 based implementation and the ABS GSIM based implementation when working down from Concept in the direction of Represented Variable / Data Element.

    Notes on comparing AIM Cascade from Concept to RV with AIHW METeOR approach.docx

    So if we set aside Variables as Concepts (the direct sub-class relationship that I didn't circle) they have two relationships to Concept, one via the Unit Type they measure and one via the "Concept as measure" that doesn't get its own name in GSIM but would be "Property" in 11179.

    11179 defines "characteristic" as an abstraction of "property".

    Where I am going here is that if we have an RV "Income of Income Unit" (where "Income Unit" is a Unit Type and 1 Household comprises 1..* Income Units and each Income Unit comprises 1..* persons) then the Value Domain is the permissible values for Income (the relevant "Concept as measure" or "Property") and not so much to do with the Unit Type.

    For example a different Variable "Income of Household" could use the same "Concept as measure", and be a candidate for using the same Value Domain, even though the Unit Type is different.

    This is why AIHW go further and recognise Generic Data Elements in their Registry to support, where applicable, measuring the same thing (Property) the same way (same Value Domain) from different "types of things" (different Unit Types / Object Classes). 

    It would be a bigger change to give "Concept as Measure" an identify in its own right in GSIM - so it may not be viable - but by swapping (using AIHW speak) FROM relating Value Domains to permissible values for an "Attribute" To relating Value Domains permissible values for (in effect, as Variable) "Entity + Attribute" we are being a little less specific and direct.