Describing variables at distribution level in StatDCAT-AP

[ChLaaboudi]

In HealthDCAT-AP, variables can be described at the distribution level, following the approach of CSV on the Web, where variable descriptions are attached to a dcat:Distribution.

In StatDCAT-AP, variables (e.g. dimensions) are instead described at the dataset level, using properties such as stat:dimension attached to dcat:Dataset.

These two approaches reflect different modelling choices:

• Dataset-level description (StatDCAT-AP) emphasises the conceptual structure of the statistical dataset.
• Distribution-level description (HealthDCAT-AP / CSVW) focuses on the structure of a specific file or representation.

Would it be relevant for StatDCAT-AP to also allow the description of variables at the distribution level, in addition to the dataset-level approach?
This could be useful in cases where the variables depend on the structure of a particular distribution (e.g. CSV files with specific columns).

[jakubklimek]

Could we get examples? DCAT Distributions of one dataset should represent the same contents - Having different varibales in different distributions would defy this, no?