Measure missing in Dataset annotations

[jakubklimek]

Currently in StatDCAT-AP, we can annotate a dataset with:

• dimensions
• attributes
• units of measurement

In a DCV DSD, we specify:

• dimensions
• attributes
• measures <- this is missing on dcat:Dataset level
• units of measurement as a special case of attribute (stat:statUnitMeasure)

The proposal is to also add stat:measure to Dataset to support representation of WHAT is measured in addition to the units of measurement used, other attributes, and the dimensions.

[ChLaaboudi]

The concept of stat:measure corresponds to the notion of Conceptual Variable in GSIM, defined as the combination of a Characteristic and a Statistical Unit — i.e. the use of a Concept as a characteristic of a Unit Type that is intended to be observed.

Conceptual variable	Characteristic	Statistical Unit
Employment status of a person	Employment status	Person
Gross revenue of an enterprise	Gross revenue	Enterprise
Age of a person	Age	Person
Number of dwellings in a municipality	Count of dwellings	Municipality

Example from #9

Conceptual variable	Characteristic	Statistical Unit
Enterprises do not use AI technologies because the costs seem too high	Barrier to AI adoption: perceived cost too high	Enterprise
Enterprises' AI technologies were commercial software or systems modified by own employees	Type of AI technology acquisition: commercial software modified in-house	Enterprise

[ChLaaboudi]

@jakubklimek
Following the discussion, would it be worthwhile to also propose a new stat:statisticalUnit property to describe the Statistical Unit — i.e. the kind of entity being observed in a dataset (e.g. person, enterprise, country)?
This would complement stat:measure and also distinct from the existing stat:statUnitMeasure, which captures the unit of measurement rather than the observed entity type.

[ChLaaboudi]

Following the discussion, would it be worthwhile to also propose a new stat:statisticalUnit property to describe the Statistical Unit — i.e. the kind of entity being observed in a dataset (e.g. person, enterprise, country)?
This would complement stat:measure and also distinct from the existing stat:statUnitMeasure, which captures the unit of measurement rather than the observed entity type.

The range of the property is a rdf:Resource (with a rdf:label or skos:prefLabel)

[jakubklimek]

@ChLaaboudi Actually, it would be a qb:MeasureProperty. If RDF DCV distribution would be present it would be the value of qb:measure in a DSD, e.g. eg:lifeExpectancy in:

eg:dsd-le a qb:DataStructureDefinition;
    # The dimensions
    qb:component [ qb:dimension eg:refArea;         qb:order 1 ];
    qb:component [ qb:dimension eg:refPeriod;       qb:order 2 ];
    qb:component [ qb:dimension sdmx-dimension:sex; qb:order 3 ];
    # The measure(s)
    qb:component [ qb:measure eg:lifeExpectancy];
    # The attributes
    qb:component [ qb:attribute sdmx-attribute:unitMeasure; 
                   qb:componentRequired "true"^^xsd:boolean;
                   qb:componentAttachment qb:DataSet; ] .

[jakubklimek]

Related issue from 2017:

STATDCAT currently provides properties to associate attributes and dimensions with datasets. It also allows the unit of measurement for the measure dimension to be specified. But it doesn't allow the measure itself to be referenced.

A DataCube DataStructureDefinition allows attributes, measures and dimensions to all be associated with a dataset. Surely STAT DCAT should do similarly?

I'm also wondering about why the approach was taken to define new properties, e.g. stat:attribute, rather than allow a qb:DataStructureDefinition to be associated with a dcat:Dataset. This would be as expressive and would only require a single new property.

Any insight appreciated!

Thanks,

L.

[ChLaaboudi]

Actually, it would be a qb:MeasureProperty. If RDF DCV distribution would be present it would be the value of qb:measure in a DSD, e.g. eg:lifeExpectancy in:

eg:dsd-le a qb:DataStructureDefinition;
# The dimensions
qb:component [ qb:dimension eg:refArea; qb:order 1 ];
qb:component [ qb:dimension eg:refPeriod; qb:order 2 ];
qb:component [ qb:dimension sdmx-dimension:sex; qb:order 3 ];
# The measure(s)
qb:component [ qb:measure eg:lifeExpectancy];
# The attributes
qb:component [ qb:attribute sdmx-attribute:unitMeasure;
qb:componentRequired "true"^^xsd:boolean;
qb:componentAttachment qb:DataSet; ] .

Actually, it would be a qb:MeasureProperty. If RDF DCV distribution would be present it would be the rdfs:label value of qb:measure in a DSD, e.g. eg:lifeExpectancy (label:life expectancy) in:

eg:dsd-le a qb:DataStructureDefinition;
    # The dimensions
    qb:component [ qb:dimension eg:refArea;         qb:order 1 ];
    qb:component [ qb:dimension eg:refPeriod;       qb:order 2 ];
    qb:component [ qb:dimension sdmx-dimension:sex; qb:order 3 ];
    # The measure(s)
    qb:component [ qb:measure eg:lifeExpectancy];
    # The attributes
    qb:component [ qb:attribute sdmx-attribute:unitMeasure; 
                   qb:componentRequired "true"^^xsd:boolean;
             qb:componentAttachment qb:DataSet; ] .

eg:lifeExpectancy  a rdf:Property, qb:MeasureProperty;
    rdfs:label "life expectancy"@en;
    rdfs:subPropertyOf sdmx-measure:obsValue;
    rdfs:range xsd:decimal .

[ChLaaboudi]

A DataCube DataStructureDefinition allows attributes, measures and dimensions to all be associated with a dataset. Surely STAT DCAT should do similarly?

I'm also wondering about why the approach was taken to define new properties, e.g. stat:attribute, rather than allow a qb:DataStructureDefinition to be associated with a dcat:Dataset. This would be as expressive and would only require a single new property.

Keeping the structural description at the distribution level is relevant with the proposal in issue #25.

Linking a dcat:Distribution to a qb:DataStructureDefinition via dct:conformsTo could provide the structural description of a statistical dataset in a way that reuses the existing RDF Data Cube vocabulary:

DCAT-AP & StatDCAT-AP

xyz a dcat:Dataset ;
    dcat:distribution :xyz_distribution .

:xyz_distribution a dcat:Distribution ;
    dct:conformsTo eg:dsd1 .

DCV

eg:dsd1 a qb:DataStructureDefinition ;
    qb:component
        [ qb:dimension sdmx-dimension:refTime ; ],
        [ qb:measure eg-measure:quantity ; ],
        [ qb:measure eg-measure:weight ; ] .

[jakubklimek]

@ChLaaboudi It is, however, not aligned with the conclusion in #21 (comment), where DSD is aligned more with a Dataset.

[ChLaaboudi]

@ChLaaboudi It is, however, not aligned with the conclusion in #21 (comment), where DSD is aligned more with a Dataset.

How to deal with a dataset with two data structures? One SDMX/XML, one DCV/RDF?