Issue #2-6: Change to relationship between Variable and Unit Type

(Feedback from Australia)

This change was missed by the ABS previously.

The title of Issue #51 is about the relationship between Unit Types and Populations but the action taken in the comment was to change the cardinality of the relationship between Variables and Unit Types.

The ABS implementation of both Variables and Unit Types is strongly aligned with the GSIM 1.1 (and 1.0) modelling.  That modelling has each Variable associated with a single Unit Type.

Even if the proposed change were considered desirable in principle, it would be an example of a change that would impact existing ABS Data Architecture so broadly that cost and risk of implementation would be unacceptably high given an apparent lack of strong practical benefits.

It is not clear, however, that the proposed change is conceptually desirable.  

The change would, for example, reduce the alignment between GSIM and reasonably commonly implemented (including beyond Official Statistics) ISO 11179 Part 3 metamodel.

• GSIM:Variable is analogous to ISO11179:DataElementConcept
• GSIM:UnitType is analogous to ISO11179:ObjectClass.
• ISO11179 has DataElementConcepts related (optionally) to a single ObjectClass.

ABS uses Unit Type as a fundamental building block for statistics.  They can be bundled together flexibly into scoped Populations (or GSIM1.5 Universes) so it is possible to define a lot from a small set of building blocks.

In the cases given on the wiki, ABS would use the Unit Type "Person" in both cases, with qualification of the Variable showing in the first case that what was being measures was the "Country of Birth of the Mother of the Person" as opposed to the "Country of Birth of the Person". 

Our current approach makes it easy to, eg, define a standard set of core demographic variables that can be applied to any Person (and a separate set to any Business) without needing to update the standard Variable each time a new "sub Unit Type" is identified as relevant to a specific Statistical Program (SP).  An SP can add a new qualification to the Variable and/or associated Population/Universe rather than adding a new "sub Unit Type".

It would be helpful to identify any compelling benefit of the change proposed in response to Item #51 compared with alternatives for accurately describing data.  This is particularly the case given the change moves GSIM out of alignment with relevant reference standards such as ISO 11179.