I wonder if our current generic attribute modelling is too narrow. The two modelling ways (quantitative = float+unit and qualitative = string) maps onto ratio/interval vs. nominal, but leaves ordinal unaddressed. Ordinal needs (e.g. for Mohs scale or Beaufort scale):
- A numeric or coded value (preserving ordering information)
- No unit (the scale is self-contained and arbitrary)
| Class |
Scale type |
value range |
unit |
class_uri |
| QuantitativeAttribute |
Ratio / Interval |
float |
required/rec. |
qudt:Quantity |
| QualitativeAttribute |
Nominal (string) |
string |
none |
prov:Entity |
| (proposed) CategoricalAttribute |
Nominal (IRI) |
DefinedTerm |
none |
qudt:EnumeratedValue |
Currently such a class may be added in the domain profile(s). However, since a CategorialAttribute class appears to be quite general, it may deserve to be added to DCAT-AP+.
According to levels of measurement this additional attribute would cover both, categorial and ordinal types (values).
I wonder if our current generic attribute modelling is too narrow. The two modelling ways (quantitative = float+unit and qualitative = string) maps onto ratio/interval vs. nominal, but leaves ordinal unaddressed. Ordinal needs (e.g. for Mohs scale or Beaufort scale):
Currently such a class may be added in the domain profile(s). However, since a CategorialAttribute class appears to be quite general, it may deserve to be added to DCAT-AP+.
According to levels of measurement this additional attribute would cover both, categorial and ordinal types (values).