value domain for descriptor variable

[smrgeoinfo]

From: A. G. ilg21@yahoo.com
Sent: Tuesday, December 2, 2025 11:37 AM

One question comes up pretty quickly. For Wide data Structures, we use SKOS to encode substantive value domains. Here, we are using the list of variable ids. I think we should probably be consistent here, unless we expect the ids to ap[pear in the data, which is typically not the case. The rule should be "when representing variables, use SKOS to describe the enumerations." This makes sense to me because categories - the meanings of the codes - are concepts, and their notations provide the symbols found in the data set. This is unambiguous

[smrgeoinfo]

I had to wrestle with how to represent the value domain for the descriptor variable. For the structure description to be useful, there has to be a mapping from the strings in the descriptorVariable field to the instance variable. — 'Characteristic' or  'CharacteristicName' in the example dataset.  Makes sense that these would be concepts, and I figured that since InstanceVariable inherits from cdi:Concept in the model that could work.  The trick is that the client using the metadata has to know that the cdi: name (or prefLabel if we use skos) is the string that shows up in the descriptor field in the data.

Also I thought the idea is that the values of the descriptor identify variables that are specified in that row/record, so defining the descriptor value domain with instance variables makes sense...