Motivation

This paper introduces COOS, the Core Ontology for Official Statistics. COOS main purpose is to serve as an integration model for the core set of ModernStats standards backed by elements of well-known standard vocabularies. These ModernStats standards, mostly developed under the auspices of the UNECE High-Level Group for the Modernisation of Official Statistics, include the Generic Statistical Business Process Model (GSBPM), the Generic Activity Model for Statistical Organisations (GAMSO), the Generic Statistical Information Model (GSIM), and the Common Statistical Data Architecture (CSDA).

As more statistical offices are turning to semantic standards to formalize their data and metadata, it became necessary to establish common foundations on which the different standards can develop in a coherent way using a formal framework that allows interoperability, machine-actionability and globally unique identification.

ModernStats standards have been developed independently over the course of more than a decade by a diverse group of specialists with different viewpoints, stakeholders and ideas. This created misalignments and impedance mismatches between the underlying models that should otherwise work well together and complement each other: Information objects (GSIM) describe the data and metadata necessary to produce statistics where capabilities (CSDA) are the essential building blocks enabling activities (GAMSO) to be implemented via business processes (GSBPM).

COOS defines a conceptual integration framework to provide semantic coherence across these models based on a common vocabulary of terms, definitions and a well-defined set of inter- and intra-model relationships formalized in RDF/OWL using standards vocabularies, e.g. SKOS, PROV, DCAT, DC, ORG, etc. COOS provides a powerful mechanism to describe complex aspects of statistical production to support business discussions and technical solution implementations.

Model management is a big part of the standards integration story: the underlying models evolve and to maintain alignment the COOS needs to evolve with them. COOS includes an initial governance framework complementing each model’s own governance processes. This governance framework includes a core set of principles and a process for managing change.

Background

ModernStats standards consist of conceptual models and supporting documentation developed for the official statistics domain describing information entities and capabilities together with business processes and activities. The essence of this core set of standards is described below.

GSBPM provides a framework to describe the building blocks of statistical production in terms of sub-processes. Its main goal is to help statistical organizations standardize their statistical production processes. It was the first ModernStats model to be published, back in 2008, and has been widely used by national and international statistical agencies since then.

GAMSO provides a framework to describe the building blocks of statistical production in terms of activities. It complements the GSBPM in two ways: (i) by covering areas beyond the scope of GSBPM, and (ii) by providing a business capability view of statistical production itself.

GSIM complements both GSBPM and GAMSO by providing a catalogue of information objects to describe statistical data and metadata. It functions as a reference framework consisting of a set of standardised information objects to be used in statistical production.

CSDA provides a capability framework cataloguing the major abilities a statistical organization has to use, produce, share and manage data and metadata. CSDA integrates with the GSBPM and GAMSO by enabling processes and activities related to the lifecycle management of GSIM information objects.

Overview

The ontology provides classes and properties to allow statistical activities to be described, as well as the organizations that undertake them, the inputs they use and the ouputs they produce.

All activities undertaken by organizations that produce Official Statistics may be represented using the ontology. A distinction is made between activities directly linked to the production of statistical content, and activities required to perform those. Several classes of activities are also introduced to allow different levels of granularity to be expressed. The ontology also provides instances of GAMSO/GSBPM's activities and the possibility to map these to user defined activities. See Section 3. Activities.

In the context of the ontology, organizations are distinguished by whether they produce Official Statistics at national or international level. See Section 4. Organizations.

In the course of a statistical production process, different kinds of information objects flow from one activity to the other. The ontology focuses on the core information object “DataSet”. See Section 5. Products.

COOS Documents

In addition to the present specification, the COOS family of documents includes:

• The formal vocabulary expressed with the Terse RDF Triple language (Turtle) [[turtle]] syntax.
• The governance document which describes the principles and processes used for the evolution of COOS.
• The URI policy which describes the principles used for the naming and identification of the COOS artifacts.

If there are any ambiguities or contradictions between the documents, the formal vocabulary will be authoritative.

Base GSBPM model

The GSBPM comprises three levels:

• Level 0, the statistical business process;
• Level 1, the phases of the statistical business process;
• Level 2, the sub-processes within each phase.

A first approach to OWL modelling of statistical activities based on the GSBPM is presented in [[GSBPM-LM]].

In COOS, a sub-class of skos:Concept named ActivityCategory is created. Each GSBPM phase and sub-process is defined as an instance of ActivityCategory and included (with a skos:inScheme relation) in a skos:ConceptScheme named GSBPM: the scheme of all the GSBPM phase and sub-process categories.

In addition, the statisticalProductionProcess individual, representing the statistical production process as a whole, is also declared as an instance of ActivityCategory and included in the GSBPM concept scheme.

The links between the different individuals are represented by the usual SKOS properties: all the sub-processes of a phase have a skos:broader link to the phase, and each phase has a skos:broader link to the statisticalProductionProcess individual.

				<http://id.unece.org/activities/phase/2> a coos:ActivityCategory ;
					skos:notation "2" ;
					skos:prefLabel "Design"@en ;
					skos:narrower <http://id.unece.org/activities/subProcess/2.3> .

				<http://id.unece.org/activities/subProcess/2.3> a coos:ActivityCategory ;
					skos:notation "2.3" ;
					skos:prefLabel "Design collection"@en ;
					skos:broader <http://id.unece.org/activities/phase/2> .
				

To materialize a clear distinction between the model itself and the statistical process which is modeled, the ontology defines classes that inherit from prov:Activity. Using PROV allows to describe an activity that really happens at a certain time, to specify who undertakes the activity and what entities it consumes or produces.

Basically, three classes are defined in this ontology: ProductionActivity, Phase and SubProcess.

ProductionActivity is a sub-class of prov:Activity. Instances of this class are intended to be larger than a GSBPM phase. If they correspond to a statistical production process they are typed using the dcterms:type property with the statisticalProductionProcess individual.

				ex:produce-study-2021 a coos:ProductionActivity ;
					rdfs:label "Produce study on legal population 2021"@en .
					
				ex:produce-census-2021 a coos:ProductionActivity ;
					rdfs:label "Produce census 2021"@en ;
					dcterms:type coos:statisticalProductionProcess .
				

Phase and SubProcess are both sub-classes of ProductionActivity. A link to the GSBPM taxonomy using the property dcterms:type is made on instances of these classes to formalize to what part of the GSBPM the activity that happened corresponds. Additionally, the idea that each phase is made up of its different sub-processes can be represented by using the Dublin Core dcterms:hasPart and dcterms:isPartOf properties.

				ex:collect-census-2021 a coos:Phase ;
					rdfs:label "Collect census 2021"@en ;
					dcterms:type <http://id.unece.org/activities/phase/6> .

				 ex:run-collection-census-2021 a coos:SubProcess ;
					rdfs:label "Run collection of census 2021"@en ;
					dcterms:isPartOf ex:collect-census-2021 ;
					dcterms:type <http://id.unece.org/activities/subProcess/4.3> .
				

[[[#image-gsbpm]]] summarizes the vocabulary terms defined so far in this document.

In addition the ontology defines three individuals of ActivityCategory to model the types of overarching activities provided by GSBPM the “Quality Management”, the “Data Management” and the “Metadata Management”.

				<http://id.unece.org/activities/overarchingActivity/0.2> a coos:ActivityCategory ;
					skos:notation "0.2"	;
					skos:prefLabel "Metadata Management"@en .
				

Adding GAMSO

Introducing GAMSO in this framework is not very difficult. The terminology used in the model shows that the main concepts are activity and activity areas. GAMSO describes two kinds of activities: those that are of statistical nature, for example “Manage Statistical Methodology”, and those that are in support of statistical activities, like “Manage Finances”. To account for this distinction, COOS defines two different classes: StatisticalActivity and Activity. Activity is a subclass of prov:Activity and a superclass of StatisticalActivity, which in turn is a superclass of the ProductionActivity introduced previously.

The GAMSO activity areas are not really activities by themselves, but rather “boxes” used to classify the activities. This makes them instances of skos:Concept to represent this “taxonomic” nature, or more precisely of a specific ActivityArea class, which is a sub-class of skos:Concept.

As with GSBPM, COOS defines individuals for each GAMSO activity and activity area, and the model itself is a skos:ConceptScheme. At the top of the scheme, an instance of ActivityCategory represents the whole activity of statistical organizations and is named officialStatistics. All four activity areas have skos:broader links to it and skos:narrower links to the categories of activity they contain.

				<http://id.unece.org/activities/activityArea/1> a coos:ActivityArea ;
					skos:notation "1" ;
					skos:prefLabel "Strategy and Leadership"@en .

				<http://id.unece.org/activities/activity/2.3> a coos:ActivityCategory ;
					skos:notation "2.3" ;
					skos:prefLabel "Monitor Capability Improvements"@en .
				

In addition, COOS defines the OverarchingActivity class, which represents a supporting activity that applies to all production phases, e.g. metadata management, data management, statistical methodology management, etc. These activities correspond to the overarching processes in the GSBPM and the statistical activities in Corporate Support in GAMSO. These activities support the execution of phases and sub-processes effectively. When harmonized and standardized, overarching activities enable the efficient execution of the entire statistical production process.

Finally, two properties are needed to describe how an Activity classified with GAMSO relates to a ProductionActivity classified with GSBPM:

• supports: an Activity helps to perform a ProductionActivity, e.g. a sub-process;
• uses: conversely, a statistical production activity will require one or several activities in order to function.

Properties uses and supports are inverses of each other. They represent weaker (looser) forms of dependencies than partitive relationships: uses/supports are about function whereas partitive relations are about constitution. An obvious example of uses/supports relation is included in the ontology for reference:

				ex:quality-management a coos:OverarchingActivity ;
					skos:prefLabel "Quality Management"@en ;
					dcterms:type <http://id.unece.org/activities/overarchingActivity/0.1> ;
					coos:uses ex:manage-quality .

				ex:manage-quality a coos:Activity ;
					skos:prefLabel "Manage Quality"@en ;
					dcterms:type <http://id.unece.org/activities/activity/3.3> ;
					coos:supports ex:quality-management .
				

[[[#image-gamso]]] gives an overview of the COOS elements defined so far in this document.

Note:

In COOS, the relation between the “Production” ActivityArea of GAMSO and the GSBPM as a concept scheme is expressed by a simple pair of rdfs:seeAlso properties. No GAMSO ActivityCategory is defined in the “Production” area, that would be redundant with the “Statistical Production Process” individual, but a skos:exactMatch relation is made between the GAMSO activity area and the GSBPM individual. It is possible that GAMSO and GSBPM evolve in the future towards a totally integrated view. In such an approach, all GSBPM and GAMSO artefacts would go into a single skos:ConceptScheme and the GSBPM would not be a concept scheme anymore but rather a skos:Collection.

Adding GSIM and CSDA

Adding the Generic Statistical Information Model (GSIM) in this framework allows us to refine StatisticalActivity even further. In parallel to ProductionActivity, two additional sub-classes of StatisticalActivity can be defined:

• StatisticalProgram: it is essentially a set of activities carried out to produce statistics. These statistics are about the set of units in scope for the program, e.g. “All persons with a university degree”, within a given subject field, e.g. income statistics, tourism, etc.
• StatisticalProgramCycle: statistical program activities are often repeated over time in iterations called cycles. A StatisticalProgramCycle is one of those iterations for a specific time and geography. To link a cycle to the corresponding program, it is recommended to use `dcterms:isPartOf`. Nevertheless, this property does not convey aspects like temporality, so a more specific property might be defined in a future version of COOS.

Another dimension to consider is information capabilities. A capability is an ability a statistical organization possesses to undertake a specific activity. It is achieved through the integration of all relevant capability elements (e.g. methods, processes, standards and frameworks, IT systems and people skills). The notion of capability is often used in enterprise architecture approaches like TOGAF (The Open Group Architecture Framework) or, in the statistical domain, the Common Statistical Data Architecture [[CSDA]]. Capabilities can be mapped to strategic goals and objectives and provide a useful starting point to map lower level elements such as business process and functions, applications and technology assets.

Following CSDA, two types of capabilities are distinguished:

• Core capabilities: capabilities the organization needs to execute its core business, i.e. the production of statistics. They generally map to phases in GSBPM.
• Cross-cutting capabilities: capabilities used to formulate and implement the policies that the organization chooses for its internal operations. They generally map to corporate support in GAMSO and overarching processes in GSBPM.

Examples of capabilities relevant to statistical activities can be found in CSDA, but also in other references like the Global Statistical Geospatial Framework (GSGF):

				ex:csda-data-integration a coos:CoreCapability ;
					rdfs:label "Data integration"@en ;
					skos:definition "The ability to combine, link, relate and/or align different data sets in order to create an integrated information set."@en .

				ex:geostat-integration-services a coos:Capability ;
					rdfs:label "Develop and apply services for smarter statistical-geospatial integration"@en ;
					rdfs:isDefinedBy <https://www.efgs.info/geostat/> .
				

Also, the European Statistical System has published a Business Capabilities Model that defines a number of capabilities organized in three levels.

To represent the link between a capability and the activities it allows to perform, COOS defines the enables property. [[[#image-csda-ex]]] below shows an example using this property.

Note:

There is an obvious relation between capabilities and the “Capability Development” GAMSO activity area, but this link is not formally modeled in COOS.

More detailed activities

The previous definitions stay in the framework of the GAMSO/GSBPM/GSIM/CSDA, but more precise notions will be needed by the statistical organizations for the more detailed modelling of their statistical activities. COOS users can adopt different names for these more specific activities that are scoped by a given GSBPM sub-process or overarching activity, but COOS defines the generic Task class as a common term for better interoperability. Tasks can be more or less granular and form hierarchies.

				ex:calculate-mean-wages a coos:Task ;
					rdfs:label "Calculate mean wages"@en .

				ex:produce-study prov:wasInformedBy ex:calculate-mean-wages .
				

[[[#image-ex1]]] shows how PROV can be used to represent links between activities: here it is supposed that the “Produce study” activity uses the results of the “Calculate mean wages” task.

As mentioned previously, the Task class can be used by statistical organizations to create instances for their own needs. Two statistical organizations can create their own "Calculate mean wages" task, and only their labels would indicate that they do similar things. In certain cases however, it could be useful to define standard sub-classes of Task for some widely used and specific types of tasks, for example record linkage or hot-deck imputation. This would allow for example to refer to relevant methodology.