DDI-RDF Discovery Vocabulary

A vocabulary for publishing metadata about data sets (research and survey data) into the Web of Linked Data

W3C Document

Editors:
Thomas Hartmann (formerly Bosch)
Richard Cyganiak (TopQuadrant)
Joachim Wackerow (GESIS - Leibniz Institute for the Social Sciences, Germany)
Benjamin Zapilko (GESIS - Leibniz Institute for the Social Sciences, Germany)
Authors:
Thomas Hartmann (formerly Bosch)
Sarven Capadisli (Enterprise Information Systems (EIS))
Franck Cotton (INSEE - Institut National de la Statistique et des Études Économiques, France)
Richard Cyganiak (TopQuadrant)
Arofan Gregory (ODaF - Open Data Foundation, USA)
Benedikt Kämpgen (Institute of Applied Informatics and Formal Description Methods (AIFB), Germany)
Olof Olsson (SND - Swedish National Data Service)
Heiko Paulheim (Data and Web Science Group, University of Mannheim, Germany)
Joachim Wackerow (GESIS - Leibniz Institute for the Social Sciences, Germany)
Benjamin Zapilko (GESIS - Leibniz Institute for the Social Sciences, Germany)

This document is also available in this non-normative format: PDF


Abstract

This specification defines the DDI-RDF Discovery Vocabulary (Disco), an RDF Schema vocabulary that enables discovery of research and survey data on the Web. It is based on DDI (Data Documentation Initiative) XML formats.

Status of This Document

This document is merely a W3C-internal document. It has no official standing of any kind and does not represent consensus of the W3C Membership.

The DDI-RDF Discovery Vocabulary is a draft specification of the DDI Alliance.

This specification is produced by the subgroup on Disco (chair Joachim Wackerow) of the RDF Vocabularies Working Group, a working group at the DDI Alliance.

Resources:

1. List of Figures

2. Introduction

The namespace for all terms in this ontology is: http://rdf-vocabulary.ddialliance.org/discovery#".

Normative formats of the DDI-RDF Discovery Vocabulary specification are

There is also a non-canonical RDF/XML version of the Turtle file.

Open issues are discussed on the issue tracker: open issues.

A detailed overview of the Disco vocabulary is available as LODE view or a web view using the web application Web-based Visualization of Ontologies.

For a detailed explanation of DDI terms please refer to section 2.

2.1 Scope and Purpose

This specification is designed to support the discovery of microdata sets and related metadata using RDF technologies in the Web of Linked Data. Many archives and other organizations have large amounts of data, sometimes publically available, but often confidential in nature, requiring applications for access. Many such organizations use the Data Documentation Initiative standard, which is a proven and highly detailed XML metadata format for describing rectangular data sets of this type. This vocabulary makes use of the DDI specification to create a simplified version of this model for the discovery of data files.

The data holdings of data archives are often collected by researchers, and only afterwards disseminated by archives. Other data-producing organizations such as research centers and statistical agencies are also increasingly interested in the DDI standards for documenting their own microdata. In general terms, most DDI metadata describes data sets for the social, behavioural, and economic sciences. This data is fairly consistent in format, consisting of rectangular data files with columns containing variables for a set of cases, contained in the rows. It is often collected by survey, although in some cases may come from administrative sources, sensors, or registers.

This vocabulary is intended not only for use by the research data community, but also by any others needing an RDF vocabulary for describing this type of rectangular data. This vocabulary will provide a useful model for describing some of the data sets now being published by open government initiatives, by providing a rich metadata structure for them. While the data sets may be available (typically as CSV files) the metadata which accompanies them is not necessarily coherent, making the discovery of these data sets difficult. This vocabulary would help to overcome this difficulty by allowing for the creation of standard queries to programmatically identify data sets, whether made available by government or held within a data archive.

Disco could be used to discover datasets by searching for specific questions, topics, and geographical coverage. Depending on the complexity of the search respectively of the data portal, parts of Disco could be used, the complete Disco, or Disco together with related vocabularies. The document [Scenarios] by Vompras, Gregory, Bosch, Capadisli, and Wackerow describes typical use cases for the applicability of the DDI-RDF Discovery vocabulary. In the Section Use Cases and Example Queries of the Appendix additional discovery use cases are illustrated by several SPARQL queries.

Statistical domain experts (core members of the DDI Alliance Technical Implementation Committee, representatives of national statistical institutes, national data archives) and Linked Open Data community members have selected the DDI elements which are seen as most important to solve problems associated with use cases in the area of data discovery. Section 2 gives an overview of the conceptual model. More detailed descriptions of all the properties are given in the specification and two conference papers [Linked-Statistical-Data] [DDI-RDF-Discovery-Vocabulary]. Disco is intended to provide means to describe microdata by essential metadata for the discovery purpose. Existing DDI-XML instances can be transformed into this RDF format and therefore exposed as Linked Data. The vice-versa process is not intended, as we have defined Disco components and reused components of other RDF vocabularies which make only sense in the Linked Data field.

2.2 About DDI

The Data Documentation Initiative standards are produced and maintained by a member-based consortium of global scope, the DDI Alliance. Housed currently at the Interuniversity Consortium for Political and Social Research (ICPSR) at the University of Michigan, there are currently more than 30 member institutions. The standards have been under development for more than ten years, and are in widespread use among data archives and libraries, producers of research data, secure data centers, and statistical agencies.

There are two major versions of DDI (both serialied in XML format): the “Codebook” version, which allows for holding general information about a study, along with its data dictionary; and the “Lifecycle” version of DDI, which allows for the description of more complex multi-wave studies, throughout the data lifecycle, from study conception through data collection and processing.

This vocabulary contains a selection of the major types of metadata defined by these two versions in a highly simplified form, for the purposes of discovery. The XML Codebook and Lifecycle versions of DDI are very broad: these standards contain hundreds of metadata elements, providing enough information to programmatically work with the data files for such functions as the automatic creation of databases, and transformations between statistical packages. DDI in both versions is generally used to describe data found in ASCII files, whether positional files with fixed-width fields or files using a delimited format such as CSV.

It is difficult to claim that there is a single agreed conceptual model for describing research data in the social, behavioural, and economic sciences—there is a wide range of models and terms. However, the issues faced in this area have been the subject of discussion within the DDI community for many years, and the DDI model represents the best consensus which exists today. As such, it gives us a good basis for creating a vocabulary which will be recognizable to researchers familiar with this type of data.

2.3 Relationship to Data Cube, DCAT and XKOS

The Discovery Vocabulary (Disco) is aligned to several other metadata vocabularies used in the RDF community. Disco is designed to be used in conjunction with other vocaularies.

The Data Catalog Vocabulary (DCAT) is a W3C standard for describing catalogs of datasets, and we map to it in two places: Our LogicalDataSet is a subclass of DCAT’s Dataset, and our DataFile is a subclass of DCAT’s Distribution. DCAT makes few assumptions about the kind of datasets being described, and focuses on general metadata about the datasets (mostly using Dublin Core), and on different ways of distributing and accessing the dataset, including availability of the dataset in multiple formats. Combining terms from both DCAT and the Discovery Vocabulary can be useful for a number of reasons:

DCAT is richer for the description of collections and catalogue. Disco supports richer descriptions of groups of datasets or individual datasets. In this spec, some of our examples are partially based on DCAT (and we will indicate when this is the case).

The Data Cube vocabulary is a W3C standard for representing data cubes, that is, multidimensional aggregate data. Data cubes are often generated by tabulating or aggregating record-level datasets. For example, if an observation in a census data cube indicates the population of a certain age group in a certain region is 12345, then this fact was obtained by aggregating that number of individual records from a record-level (or “microdata”) dataset. The Discovery Vocabulary contains a property “aggregation” (pointing from a Disco data set to a Data Cube dataset) that indicates that a Cube dataset was derived by tabulating a record-level dataset.

Data Cube provides for the description of the structure of such cubes, but also for the representation of the cube data itself, that is, the observations that make up the cube dataset. This is not the case for the Discovery Vocabulary, which only describes the structure of a dataset, but is not concerned with representing the actual data in it. The actual data is assumed to sit in a data file (e.g., a CSV file, or in a proprietary statistics package file format) that is not represented in RDF.

The interplay of Data Cube and Disco needs further exploration regarding the relationship of aggregate data, aggregation methods, and the underlying microdata. The goal would be to drill down to the related microdata based on a search resulting in aggregate data. On the one hand aggregate data are often easily available and gives a quick overview. On the other hand microdata enable more detailed analyses.

The use of formal statistical classifications is very common in research data sets—these are treated in our vocabulary as SKOS concepts, but in some cases those working with formal statistical classifications may desire more expressive capability than SKOS provides. To support such users, the DDI Alliance also publishes XKOS, a vocabulary which extends SKOS to allow for a more complete description of such classifications. While the use of XKOS is not required by this vocabulary, the two are designed to work in complementary fashion.

More details on the relationship to Data Cube, DCAT and XKOS as well as to other vocabularies are provided in Section 9.

3. Overview

Figure 1 Vocabulary Overview

To understand the DDI Discovery Vocabulary, there are a few central classes, which can serve as entry points. The first of these is the Study class. A Study in our model represents the process by which a data set was generated or collected. Literal properties include information about the funding, organizational affiliation, abstract, title, version, and other such high-level information. In some cases, where data collection is cyclic or on-going, data sets may be released as a StudyGroup, where each cycle or "wave" of the data collection activity produces one or more data sets. This is typical for longitudinal studies, panel studies, and other types of "series" (to use the DDI term). In this case, a number of Study objects would be collected into a single StudyGroup.

Data sets have two representations in our model: a logical representation, which describes the contents of the data set, and a physical representation, which is a distributed file holding that data. It is possible to format data files in many different ways, even if the logical content is the same. In our model the LogicalDataSet represents the content of the file (its organization into a set of variables (Variable)). The LogicalDataSet is an extension of the dcat:DataSet class. Physical, distributed files are represented by the class DataFile (not depicted in the diagram), which is itself an extension of the dcat:Distribution.

When it comes to understanding the contents of the data set, this is done using the Variable class. Variables (Variable) provide a definition of the column in a rectangular data file, and can associate it with a particular Concept, and a Question (the Question in the Questionnaire which was used to collect the data). Variables (Variable) are related to a representation of some form, which may be a set of codes and categories (a "codelist") or may be one of other normal data types (dateTime, numeric, textual, etc.) Codes and Categories are represented using SKOS concepts and concept schemes.

Data is collected about a specific phenomenon, typically involving some target population, and focusing on the analysis of a particular type of subject. These are respectively represented by the Universe class and the AnalysisUnit class. If, for example, the adult population of Finland is being studied, the AnalysisUnit would be individuals or persons and the Universe would be the adult population of Finland. Bosch, Cyganiak, Wackerow, and Zapilko give a detailed overview of the DDI-RDF Discovery Vocabulary in a full paper written for the Dublin Core conference [Linked-Statistical-Data].

4. Real-life Example

We have a sample of a survey which has been documented using DDI XML—the 1980 Argentine National Population and Housing Census. We are using for this example the version disseminated by IPUMS, which provides internationally harmonized census data, to make it more useful for cross-border research. Thus, this data set is produced by two organizations: The Argentine National Institute of Statistics and Censuses, and the Minnesota Population Center hosted in the University of Minnesota.

To give some idea of what is contained in the metadata set, we will use some screen shots from OpenMetadata Survey Catalog, a portal which indexes the DDI files to facilitate searching, and reflects the contents in a fashion which is easy to view. Follow this link for the information about this DDI file at the OpenMetadata Survey Catalog.

Figure 2 Overview

Figure 2 shows us the overview page for this study, giving us some basic information - title, identifier for the study, data producers, year, country, and a link to the access policies. If we look at the right-hand panel, we see an outline of the metadata contents of the file, including information about the questionnaire used, sampling methodology, and data collection activities, as well as the two data files which contains detailed information about its variables.

Not all of this information is useful in a data discovery scenario—sampling and data collection methodologies are not typically indexed for searches. Information about the questionnaire is, as is detailed information about the variables contained in the files. We will look more closely at the metadata of primary interest for our discovery scenario.

Using RDF and the DDI Discovery Vocabulary, the study can also be described in triples: an instance of type of Study is given the title and the identifier; also, the two data producers are linked and further described. The year and country are described in the form of a temporal and spatial coverage of the study. Also, the topics of the study are represented. The study instance further contains an abstract. Since a study is a versionable object in DDI, we attach a version to it. A study is further described using additional information which is described in the following Example 1.
# We will use the namespace 'ddi' in all of our examples.

ddi:Study_1 a disco:Study;
    dcterms:title "National Population and Housing Census, 1980"@en;
    dcterms:identifier "ARG_1980_PHC_v01_A_IPUMS";
    dcterms:creator [
        rdfs:label "Minnesota Population Center"@en;
        skos:notation "MPC";
        org:memberOf [
            rdfs:label "University of Minnesota"@en;
        ];
    ];
    dcterms:creator [
        rdfs:label "Argentine National institute of Statistics and Censuses"@en;
    ]
	dcterms:temporal [
		a dcterms:PeriodOfTime ;
		disco:startDate "1980-10-22"^^xsd:date;
		disco:endDate "1980-10-22"^^xsd:date;
		rdfs:comment "The interviews take place on the expected census day. In
		  	      some areas the enumeration took place the following day because of
		  	      access problems due to heavy rains.";
    ];
    dcterms:spatial [
      # This is the DC-strictly compatible way to do it
      a dcterms:Location;
      rdfs:label "Argentina, national coverage"@en;
    ];
    # Only a subset of subjects mentioned in the original file
    dcterms:subject [
		skos:definition "Technical Variables -- HOUSEHOLD"@en ;
    ] ;
    dcterms:subject [
		skos:definition "Group Quarters Variables -- HOUSEHOLD"@en ;
    ] ;
	dcterms:abstract "IPUMS-International is an effort to inventory, preserve,
		         harmonize, and disseminate census microdata from around the world. The
		         project has collected the world's largest archive of publicly available
         		 census samples. The data are coded and documented consistently across
		         countries and over time to facilitate comparative research. IPUMS-
		         International makes these data available to qualified researchers free
		         of charge through a web dissemination system. The IPUMS project is a
		         collaboration of the Minnesota Population Center, National Statistical
		         Offices, and international data archives. Major funding is provided by
		         the U.S. National Science Foundation and the Demographic and Behavioral
		         Sciences Branch of the National Institute of Child Health and Human
		         Development. Additional support is provided by the University of
		         Minnesota Office of the Vice President for Research, the Minnesota
		         Population Center, and Sun Microsystems.";    

    owl:versionInfo "Version 1.0. This version contains selected variables from
		    the original census microdata plus harmonized variables from the IPUMS
		    International data base."@en; 

    disco:universe ddi:Universe_1;
    disco:instrument ddi:Questionnaire_1;
    disco:product ddi:Dataset_1;
    
    disco:analysisUnit ddi:AnalysisUnit_1;
    disco:kindOfData ddi:KindOfData_1;

    # stdyInfo/notes currently not represented. 
    disco:variable ddi:AR80A401, ddi:AR80A402, ddi:AR80A404, ddi:AR80A407, ddi:AR80A411.

While the sampling methodology may not be of great interest for those searching for data, one field within this section is: the “universe”, that is, the population being studied. Figure 3 gives us an example of this information.

Figure 3 Coverage and Universe

Thus, the study refers to a specific universe.

ddi:Universe_1 a disco:Universe;
    skos:definition "All the population in the national territory at the moment the census is carried out."@en .
Using a type of instrument - a questionnaire -, the study produced a dataset. The dataset has access rights. The dataset has a concrete data file (physical representation or distributed file) populated by certain variables.
ddi:Dataset_1 a disco:LogicalDataSet;
    disco:instrument ddi:Questionnaire_1;
    dcterms:accessRights ddi:AccessRights_1;
    disco:dataFile ddi:Datafile_1;
    disco:variable ddi:AR80A401, ddi:AR80A402, ddi:AR80A404, ddi:AR80A407, ddi:AR80A411.

ddi:AccessRights_1 a dctermsRightsStatement;
  dcterms:description "IPUMS-International distributes
	integrated microdata of individuals and households only by agreement ... 
	designed to extend this record.";
    rdfs:seeAlso <http://microdata.worldbank.org/index.php/catalog/442/accesspolicy>.

Figure 4 shows us the information about access policies, which typically is of interest to those searching for data.

Figure 4 Access Policy

The Unit of Analysis and Kind of Data further describe the study.

ddi:AnalysisUnit_1 a disco:AnalysisUnit ;
    skos:definition "Dwelling, quarter dwelling, census household, and population"@en .

ddi:KindOfData_1 a skos:Concept ;
	rdfs:label "Census/enumeration data [cen]"@en .

In some cases we may have a lot of information about the questionnaires used, and it is very common to search for data by the text of the question used to collect it. Sometimes there will be a PDF of a questionnaire, and sometimes question text may be linked to individual variables within a file. In this case, we have only a textual description of the set of forms used in the census (Figure 5).

Figure 5 Questionnaires

The following example illustrates three questions. Each question does have a text.

ddi:Questionnaire_1 a disco:Questionnaire;
    disco:question ddi:QuestionGender;
    disco:question ddi:QuestionAge;
    disco:question ddi:QuestionCitizenship.

ddi:QuestionGender a disco:Question;
    disco:questionText "2. Is the person a man or a woman? [] Man, [] Woman"@en.

ddi:QuestionAge a disco:Question;
    disco:questionText "3. What is his or her age? _ _ Mark the age in completed
		years at the date of the census for those younger than one year old mark
		00. For those younger than 10 years old, mark 01, 02, 03, etc. For those
		older than 99 years old, mark 99."@en.

ddi:QuestionCitizenship a disco:Question;
    disco:questionText "6. [Immigration status] Only for persons who have usual
		residence in Argentina and were born in another country. [Questions 6A
		and 6B asked only of persons born outside Argentina and who currently
		reside in Argentina.] B. Are you a naturalized citizen of Argentina?
		[] Yes [] No [] Unanswered"@en.

In Figure 6 we see the list of variables contained in the data file. For each of these we will also have a detailed view, showing the codes and categories used to encode the actual responses in the variables (Figure 7).

Figure 6 Variables List
Figure 7 Variable Details

Any variable has a text and is based on a variable definition.

Note

Please note that the turtle example describes the variable labels from the screenshot above and references to the related represented variable and question.

ddi:AR80A401 a disco:Variable;
    dcterms:identifier "AR80A401";
    skos:prefLabel "Sex"@en, "Sexe"@fr;
    dcterms:description "This variable indicates the person's gender."@en;
    disco:basedOn ddi:SexVD;
    disco:question ddi:QuestionGender.

ddi:AR80A402 a disco:Variable;
    dcterms:identifier "AR80A402";
    dcterms:description "This variable indicates the person's age in years."@en;
    skos:prefLabel "Age"@en, "Âge"@fr.
    disco:basedOn ddi:AgeVD;
    disco:question ddi:QuestionAge.   

ddi:AR80A407 a disco:Variable;
    dcterms:identifier "AR80A407";
    dcterms:description "This variable indicates whether or not the person is
		a naturalized citizen of Argentina."@en;
    skos:prefLabel "Citizenship"@en, "Citoyenneté"@fr;
    disco:basedOn ddi:CitizenshipVD;
    disco:question ddi:QuestionCitizenship.
Any variable definition has a representation defining the possible values of a variable. Also, a variable definition has its own universe (may be the same as the study or possibly narrower) and (DDI) concepts further describing the variable.
ddi:SexVD a disco:RepresentedVariable;
    disco:universe ddi:UniversePerson;
    disco:representation ddi:SexRepr;
    disco:concept ddi:IpumsC1;
    skos:prefLabel "Sex"@en, "Sexe"@fr;
    dcterms:description "Sex data element"@en.

ddi:SexRepr a skos:ConceptScheme, disco:Representation;
    skos:hasTopConcept ddi:SexM, ddi:SexF.

ddi:SexM a skos:Concept;
    skos:notation "1";
    skos:prefLabel "Male"@en, "Homme"@fr;
    skos:inScheme ddi:SexRepr.

ddi:SexF a skos:Concept;
    skos:notation "2";
    skos:prefLabel "Female"@en, "Femme"@fr;
    skos:inScheme ddi:SexRepr.
    
ddi:ageVD a disco:RepresentedVariable;
    disco:universe ddi:UniversePerson;
    disco:representation ddi:AgeRepr;
    disco:concept ddi:IpumsC1;
    skos:prefLabel "Age"@en, "Âge"@fr;
    dcterms:description "Age data element"@en.

ddi:AgeRepr a skos:ConceptScheme, disco:Representation;
    skos:hasTopConcept ddi:Age0, ddi:Age1, ddi:Age99.

ddi:Age0 a skos:Concept;
    skos:notation "0";
    skos:prefLabel "0";
    skos:inScheme ddi:AgeRepr.

ddi:Age1 a skos:Concept;
    skos:notation "1";
    skos:prefLabel "1";
    skos:inScheme ddi:AgeRepr.

# ...

ddi:Age99 a skos:Concept;
    skos:notation "99";
    skos:prefLabel "99";
    skos:inScheme ddi:AgeRepr.
    
ddi:CitizenshipVD a disco:RepresentedVariable;
    disco:universe ddi:UniverseNonArgentines;
    disco:representation ddi:CitizenshipRepr;
    disco:concept ddi:IpumsC2;
    skos:prefLabel "Citizenship"@en;
    dcterms:description "Citizenship data element"@en.

ddi:CitizenshipRepr a skos:ConceptScheme, disco:Representation;
    skos:hasTopConcept ddi:CYes, ddi:CNo, ddi:CUnknown, ddi:CNIU.

ddi:CYes a skos:Concept;
    skos:notation "1";
    skos:prefLabel "Yes";
    skos:inScheme ddi:CitizenshipRepr.

ddi:CNo a skos:Concept;
    skos:notation "2";
    skos:prefLabel "No";
    skos:inScheme ddi:CitizenshipRepr.

ddi:CUnknown a skos:Concept;
    skos:notation "8";
    skos:prefLabel "Unknown";
    skos:inScheme ddi:CitizenshipRepr.

ddi:CNIU a skos:Concept;
    skos:notation "9";
    skos:prefLabel "NIU (not in universe)";
    skos:inScheme ddi:CitizenshipRepr.
Any universe of a variable definition is a subset of the universe of the entire study. In our example, two questions are addressing the universe of persons, the third question is addressing a specific subset of the universe of persons.
ddi:UniversePerson a disco:Universe;
    skos:definition "All persons."@en ;
    skos:narrower ddi:Universe_1.

ddi:UniverseNonArgentines a disco:Universe;
    skos:definition "Foreign-born persons who reside in Argentina."@en ;
    skos:narrower ddi:Universe_1; 
    skos:narrower ddi:UniversePerson.

At the bottom of the screen showing the variable detail, we can see that the variable for roofing material is associated with a high-level concept, “Dwelling characteristics variables.” (Figure 8.)

In Disco, DDI concepts can be hierarchically structured

ddi:IpumsCS a skos:ConceptScheme;
    skos:hasTopConcept ddi:IpumsC1.

ddi:IpumsC1 a skos:Concept;
    skos:prefLabel "Demographic Variables - PERSON"@en, "Variables démographiques - PERSONNE"@fr;
    skos:inScheme ddi:IpumsCS.

ddi:IpumsC2 a skos:Concept;
    skos:prefLabel "Nativity and Birthplace Variables -- PERSON"@en;
    skos:inScheme ddi:IpumsCS.
The variable within a data file can be described using category statistics. In the following example, absolute and relative frequencies of the variable categories are described. This variable represents the sex of the respondent. A variable is represented by a code list containing the code, the category statistics resource is pointing to.
ddi:CatStatistics_1 a disco:CategoryStatistics;
    disco:frequency 13314444;
    disco:percentage 49.97;
    disco:statisticsCategory ddi:SexM;
    disco:statisticsDataFile ddi:Datafile_1.

ddi:CatStatistics_2 a disco:CategoryStatistics;
    disco:frequency 1336270;
    disco:statisticsCategory ddi:SexF;
    disco:statisticsDataFile ddi:Datafile_1.

Next we find some general information about the data files produced by this study (Figure 9).

Figure 9 General Data Set Information

Finally, the data file more concretely describes the actual physical file.

ddi:Datafile_1 a disco:Datafile;
    dcterms:identifier "ARG1900-P-H.dat";
    dcterms:description "Person records"@en;
    disco:caseQuantity 2667714;
    dcterms:format "ascii";
    dcterms:provenance "Minnesota Population Center"@en;
    owl:versionInfo "Version 1.0, IPUMS sample"@en;
    dcterms:spatial [
        # This is the DC-strictly compatible way to do it
        a dcterms:Location;
        rdfs:label "Argentina, national coverage"@en
    ];
    dcterms:temporal "PeriodOfTime"@en;
    dcterms:subject "To be defined"@en.

5. Studies and StudyGroups

A simple Study supports the stages of the full data lifecycle in a modular manner. A Study represents the process by which a data set was generated or collected. Literal properties include information about the funding, organizational affiliation, abstract, title, version, and other such high-level information. The key criteria for a study are: a single conceptual model (e.g. survey research concept), a single instrument (e.g. questionnaire) made up of one or more parts (ex. employer survey, worker survey), and a single logical data structure of the initial raw data (multiple data files can be created from this such as a public use microdata file or aggregate data files). In some cases, where data collection is cyclic or on-going, data sets may be released as a StudyGroup, where each cycle or "wave" of the data collection activity produces one or more data sets. This is typical for longitudinal studies, panel studies, and other types of "series" (to use the DDI term). In this case, a number of Study objects would be collected into a single StudyGroup.

Studies (Study) may be contained in at most 1 StudyGroup and groups of studies may include 0 to n studies. Studies (Study) may have 0 to n instruments (Instrument) relationships to instruments (Instrument). Particular instruments (Instrument), however, are connected with exactly 1 Study. Studies (Study) may have DataFile connections with 0 to n data files (DataFile) and data files (DataFile) must have 1 to n DataFile relationships to studies (Study). Studies (Study) are associated with 0 to n variables (Variable) using the object property Variable. On the other hand, variables (Variable) must be related to 1 to n studies (Study). Studies (Study) may have 0 to n logical data sets (LogicalDataSet) (product) and logical data sets (LogicalDataSet) must have 1 to n product relationships to studies (Study).

5.1 Coverage, References to DDI-XML Files, and Kind of Data

Figure 10 Coverage, References to DDI-XML Files, and Kind of Data

Studies (Study) or groups of studies (StudyGroup) (the union of Study and groups of studies (StudyGroup)) may have different datatype properties. Studies (Study) or groups of studies (StudyGroup) may have an abstract (dcterms:abstract), a title (dcterms:title), a subtitle (subtitle), an alternative title (dcterms:alternative), a purpose (purpose), and information about the date and the time since when the Study is publicly available (dcterms:available). Studies (Study) or groups of studies (StudyGroup) may have multiple object properties. The object properties kindOfData and dcterms:subject guide to skos:Concepts. kindOfData describes, with a string or a term from a controlled vocabulary, the kind of data documented in the logical product(s) of a Study. Examples include survey data, census/enumeration data, administrative data, measurement data, assessment data, demographic data, voting data, etc. Coverage describes the temporal, spatial and topical coverage of a study. Coverage specifies the population from which observations for a particular topic can be drawn. You can use dcterms:subject to describe the topical coverage of studies (Study) and groups of studies (StudyGroup). ddiFile to foaf:Documents which are the DDI-XML files containing further descriptions of the Study or the StudyGroup. Use dcterms:temporal for temporal coverages related to the union of studies (Study) and groups of studies (StudyGroups). For the spatial coverage use dcterms:spatial. The cardinalities of all the object properties are in both directions 0 to n. The only exception is that studies (Study) and groups of studies (StudyGroup) may have 0 or 1 kindOfData relationships to skos:Concepts.

5.2 Relationships to Agents

Figure 11 Relationships to Agents

Creators (dcterms:creator), contributors (dcterms:contributor), and publishers (dcterms:publisher) of Studies (Study) and groups of studies (StudyGroup) are foaf:Agents which are either foaf:Persons or org:Organizations whose members are foaf:Persons. Studies (Study) or groups of studies (StudyGroup) may be funded by (fundedBy) foaf:Agents. The object property fundedBy is defined as sub-property of dcterms:contributor. The cardinalities of these object properties are in both directions always 0 to n. foaf:Agents may have roles such as analyst, data modeler, programmer, and co-investigator. These roles are represented using skos:Concepts. foaf:Agents and skos:Concepts are related by disco:hadRole. Roles can be defined (skos:definition), identified (skos:notation), and described (skos:prefLabel).

5.3 Analysis Units and Universes

Universe is the total membership or population of a defined class of people, objects or events. A population is the number of statistical units sharing at least one common property which is of interest in a statistical analysis. There are two types of population, target population and survey population. A target population is the population outlined in the survey objects about which information is to be sought. A survey population (also known as the coverage of the survey) is the population from which information can be obtained in the survey. AnalysisUnit is defined as follows: The process of collecting data focuses on the analysis of a particular type of subject. If, for example, the adult population of Finland is being studied, the AnalysisUnit would be individuals or persons.

Figure 12 Study, Universe and AnalysisUnit

Studies (Study) and groups of studies (StudyGroup) must have 1 to n universes (Universe) and 1 particular Universe may be in a Universe relationship with 0 to n unions of Studies ( Study) and groups of studies (StudyGroup). Universes (Universe) are sub-classes of skos:Concepts. For universes (Universe) you can state definitions using skos:definition. The union of Study and StudyGroup may have 0 or 1 AnalysisUnit reached by the object property AnalysisUnit and a specific AnalysisUnit may be in a AnalysisUnit relationship to 0 to n studies (Study) or groups of studies (StudyGroup). AnalysisUnit is specified as a sub-class of skos:Concepts.

6. General Metadata

6.1 Identification

In DDI, a lot of entities hold particular identifiers. This can be identifiers for different versions of DDI, but also persistent identifiers for, e.g. persons or organizations, that are encoded in a particular identifier scheme, e.g. ORCID or FundRef. In general, such identifiers can be added to each entitiy in DDI-RDF, since every entity is defined as an rdfs:Resource. General metadata elements which can be used on every resource in a DDI-RDF description include:

Each Disco resource must have an identifier (see figure below). The identifier is stated using the object property adms:identifier pointing from any rdfs:Ressource to 1 to n identifiers (adms:Identifier). The class adms:Identifier can include the actual identifier itself and information on identifier scheme, its version, and its agency.

Figure 13 Identification
ddi:Study_1 a disco:Study;
  dcterms:title "National Population and Housing Census, 1980"@en;
  adms:identifier [ a adms:Identifier; 
    skos:notation "us:ddi:us.mpc:ARG_1980_PHC_v01_A_IPUMS:1"; 
    adms:schemaAgency "DDI Alliance"@en.
  ];
  dcterms:creator [
    rdfs:label "Minnesota Population Center"@en;
    skos:notation "MPC";
    adms:identifier [ a adms:Identifier; 
      skos:notation "us.mpc"; 
      adms:schemaAgency "DDI Alliance"@en.
    ];
  ].

See section 'Asset Description Metadata Schema (ADMS)' for more information about the reuse of ADMS for representing identifiers.

6.2 Versioning Information

Use of the owl:versionInfo property is recommended to indicate the version number and/or additional versioning text of entities.

Any entity can have version information. As you can see in the next UML class diagram, the property owl:versionInfo has rdfs:Resource as domain. As a consequence, each DDI object can have attached versioning information. However, the most typical cases are:

Figure 14 Versioning Information

6.4 Access Rights Statements and Licenses

Every logical dataset may have access rights statements and licensing information attached to it. For those purposes, the Dublin Core properties dcterms:accessRights and dcterms:license are used.

Access rights are defined in a dcterms:RightsStatement object, which may reference an external document stating the access rights in more detail (rdfs:seeAlso). For dcterms:RightsStatements descriptions (dcterms:description) and labels (skos:prefLabel) can be assigned:

ddi:Dataset_1 a disco:LogicalDataSet ;
  dcterms:accessRights ex:AccessRights1 .
  
ddi:AccessRights_1 a dcterms:RightStatement ;
  dcterms:description "Everybody may see access this document." ;
  rdfs:seeAlso <http://www.example.org/access.html> .

License information is captured in a dcterms:LicenseDocument, which is a subtype of dcterms:RightsStatements:

ddi:Dataset_1 a disco:LogicalDataSet ;
  dcterms:license ddi:License_1 .
  
ddi:License_1 a dcterms:LicenseDocument ;
  dcterms:description "Published under Open Content License." ;
  skos:prefLabel "OCL 1.0" ;
  rdfs:seeAlso <http://opencontent.org/opl.shtml> .
Figure 16 Access Rights Statements and Licenses

Logical data sets (LogicalDataSet) may have dcterms:accessRights relationships to dcterms:RightsStatements and dcterms:license connections with dcterms:LicenseDocument. dcterms:RightsStatements is associated with foaf:Documents using the object property rdfs:seeAlso. The multiplicities for these object properties are in any case 0 to n.

6.5 Coverage of Studies, Logical Datasets, and Data Files

Coverage comprehends the key features of the scope of the data (e.g. geographic product occupation). Studies (Study), logical datasets, and data files may have a spatial, temporal, and topical coverage. Unlike in DDI-XML, there is no dedicated Coverage type in DDI-RDF. The comprehensive description by spatial, temporal, and topical coverage is directly attached to the respective study, logical dataset, and datafile (using DCMI terms).

For spatial coverage, dcterms:spatial is used, pointing to any geographic location (dcterms:Location):

ddi:Study_1 dcterms:spatial <http://sws.geonames.org/2921044/> .

In this example, Geonames is used to refer to a spatial region, in this case, the country Germany. Geonames provides URIs for continents, countries, regions, and cities, among others, and is therefore a possible option to use for describing spatial coverage.

Figure 17 Study Coverage

For temporal coverage, dcterms:temporal is used pointing to dcterms:PeriodOfTime. For time periods, labels can be attachted ( skos:prefLabel). It is also possible to define start (startDate) and end dates (endDate). Please note that these properties are a feature at risk, since the domain is not Disco. Maintainers of the domain ontology may introduce their own properties for this purpose. A possible way to describe temporal coverage is the use of the W3C time ontology:

ddi:Study_1 dcterms:temporal [
  a time:Interval ;
  time:hasBeginning [ time:inXSDDateTime
    "2012-01-01T00:00:00+01:00"^^xsd:dateTime ];
  time:hasEnd [ time:inXSDDateTime
    "2012-01-31T23:59:59+01:00"^^xsd:dateTime ] ] .

This example describes a study that has been conducted between January 1st and January 31st.

Figure 18 LogicalDataSet Coverage

Topical coverage can be expressed using dcterms:subject. DDI-RDF foresees the use skos:Concept for the description of topical coverage:

ddi:Study_1 dcterms:subject [
  a skos:Concept ;
  skos:prefLabel "Alcohol consumption" ] .
Figure 19 DataFile Coverage

The multiplicities for each of the three object properties dcterms:subject, dcterms:temporal, and dcterms:spatial are in any case 0 to n.

6.6 Other General Dublin Core Metadata Properties

The following elements from Dublin Core may be used to describe general metadata of DDI-RDF elements (see the DC definitions for more detailed descriptions):

7. Data Sets, Data Files, and Descriptive Statistics

Data sets have two representations in our model: a logical representation, which describes the contents of the data set, and a physical representation, which is a distributed file holding that data. It is possible to format data files in many different ways, even if the logical content is the same. In our model the LogicalDataSet represents the content of the file (its organization into a set of variables (Variable)). The LogicalDataSet is an extension of the dcat:DataSet class. Physical, distributed files are represented by the class DataFile, which is itself an extension of the dcat:Distribution. DescriptiveStatistics , i.e. SummaryStatistics as well as CategoryStatistics, are associated with data files ( DataFile) by the object property statisticsDataFile. Descriptive statistics simply describe what the data shows. See also the entry on descriptive statistics in the OECD glossary of statistical terms.

Figure 20 Overview: Data Sets, Data Files, Descriptive Statistics

Logical data sets (LogicalDataSet) and data files (DataFile) are connected using the object property data files (DataFile). A specific logical data set (LogicalDataSet) may be linked to 0 to n data files (DataFile) and a particular DataFile may be connected with 0 to n logical data sets (LogicalDataSet) via DataFile. DescriptiveStatistics are accociated with data files ( DataFile) by the object property statisticsDataFile. A concrete DescriptiveStatistics object may have statisticsDataFile relationships to multiple (0 - n) data files (DataFile). Data files (DataFile), however, may have 0 to n statisticsDataFile relations to DescriptiveStatistics instances.

7.1 LogicalDataSet

Each study has a set of logical metadata (LogicalDataSet) associated with the processing of data, at the time of collection or later during cleaning, and re-coding. LogicalDataSet represents the microdata dataset.

Figure 21 LogicalDataSet

LogicalDataSet is defined as a sub-class of dcat:Dataset. You can state a title (dcterms:title) and a flag indicating if the microdata dataset is publicly available (isPublic). You can specify access rights (dcterms:accessRights) and LicenseStatements (dcterms:license) for microdata datasets. For a LogicalDataSet the three dimensions of coverage can be specified: Spatial (dcterms:spatial), temporal (dcterms:temporal), and topical (dcterms:subject). The cardinalities of the object properties dcterms:spatial, dcterms:temporal, dcteerms:subject, dcterms:accessRights, and dcterms:license are 0 to n. Microdata datasets may have Instrument associations to multiple (0 - n) instruments (Instrument) and instruments (Instrument) are connected with multiple (0 - n) logical data sets (LogicalDataSet). Each LogicalDataSet has exactly 1 Universe (Universe) and one specific Universe may be in multiple (0 - n) Universe relations to logical data sets (LogicalDataSet). Logical data sets (LogicalDataSet) may contain (variable) 0 to n variables (Variable) and variables ( Variable) must be contained in 1 to n logical data sets (LogicalDataSet). Logical data sets (LogicalDataSet) can be aggregated (aggregation) to 0 to n data sets (qb:DataSet) and data sets ( qb:DataSet) can be aggregations of 0 to n logical data sets (LogicalDataSet). At last, logical data sets (LogicalDataSet) refer to 0 to n data files (DataFile) using the object property data files (DataFile) and data files (DataFile) may be linked to 0 to n logical data sets (LogicalDataSet). The class qb:DataSet is defined in the RDF Data Cube Vocabulary. 0 to n data sets (qb:DataSet) may point to multiple (0 - n) variables (Variable) (inputVariable). Please note that this property is a feature at risk, since the domain is not Disco. Maintainers of the domain ontology may introduce their own property for this purpose. Just like there is the caseQuantity data property on DataFile, there is also the data property variableQuantity on DataFile and LogicalDataSet. This is useful to have when (1) no variable level information is available and when (2) only a stub of the RDF is requested e.g when returning basic information on a study of file, we do not need to return information on potentially hundreds or thousands of variables references or metadata.

ddi:Dataset_1 a LogicalDataSet;
    dcterms:accessRights ddi:AccessRights_1;
    disco:dataFile ddi:Datafile_1;
    disco:instrument ddi:Questionnaire_1;
    disco:variable ddi:AR80A401, ddi:AR80A402, ddi:AR80A404, ddi:AR80A407, ddi:AR80A411.

7.2 DataFile

The collected data result in the microdata represented by the DataFile. Data sets have a logical representation, which describes the contents of the data set, and a physical representation, which is a distributed file holding that data. It is possible to format data files in many different ways, even if the logical content is the same. data files (DataFile), which are also dcmitype:Datasets as well as dcat:Distributions, represents all the physical distributed data files containing the microdata datasets.

Figure 22 DataFile
ddi:Datafile_1 a disco:Datafile;
  dcterms:identifier "ARG1900-P-H.dat";
  dcterms:description "Person records"@en;
  disco:caseQuantity 2667714;
  dcterms:format "ascii";
  dcterms:provenance "Minnesota Population Center"@en;
  owl:versionInfo "Version 1.0, IPUMS sample"@en;
  dcterms:spatial [
      # This is the DC-strictly compatible way to do it
      a dcterms:Location;
      rdfs:label "Argentina, national coverage"@en
  ];
  dcterms:temporal "PeriodOfTime"@en;
  dcterms:subject "To be defined"@en.

It is possible to describe data files (DataFile) (dcterms:description). Data files (DataFile), case quantities (disco:caseQuantity) and versions (owl:versionInfo) can also be stated. Using the object property dcterms:format, data files (DataFile) formats can be defined. Data files (DataFile) must have exactly 1 dcterms:format relationship to an instance of the class dcterms:MediaTypeOrExtend which is a sub-class of skos:Concept. Specific formats can be assigned to multiple (0 - n) data files (DataFile). Provenance information can be assigned to data files (DataFile). Data files ( DataFile) may have multiple (0 - n) dcterms:provenance relationships to dcterms:ProvenanceStatements. Dcterms:ProvenanceStatements, however, may have 0 to n dcterms:provenance relations to data files (DataFile). The topical, spatial, and temporal coverage of data files (DataFile) is realized by the object properties dcterms:subject, dcterms:spatial, and dcterms:temporal, all with the cardinalities 0 to n on both sides. Just like there is the caseQuantity data property on DataFile, there is also the data property variableQuantity on DataFile and LogicalDataSet. This is useful to have when (1) no variable level information is available and when (2) only a stub of the RDF is requested e.g when returning basic information on a study of file, we do not need to return information on potentially hundreds or thousands of variables references or metadata.

7.3 DescriptiveStatistics

An overview over the microdata can be given either by the descriptive statistics or the aggregated data. DescriptiveStatistics may be minimal, maximal, mean values, and absolute and relative frequencies. qb:DataSet originates from the RDF Data Cube Vocabulary, an approach to map the SDMX information model to an ontology. A qb:DataSet represents aggregated data (also known as macrodata) such as multi-dimensional tables. Aggregated data are derived from microdata by statistics on groups, or aggregates such as counts, means, or frequencies. SummaryStatistics pointing to variables and CategoryStatistics pointing to categories and codes are both descriptive statistics.

Figure 23 DescriptiveStatistics

DescriptiveStatistics may have statisticsDataFile relations to 0 to n data files (DataFile) and data files (DataFile) may be in 0 to n statisticsDataFile relations to DescriptiveStatistics individuals. SummaryStatistics point to 0 to n variables (Variable) using the object property statisticsVariable. Variables (Variable), however, may be in 0 to n of such relationships to SummaryStatistics objects. CategoryStatistics may be connected with 0 to n skos:Concepts using the property statisticsCategory and skos:Concepts representing codes (values) and categories (value labels) may be in 0 to n of such relationships. SummaryStatistics and CategoryStatistics may have a weightedBy relation to a Variable. A statistical weight is an amount given to increase or decreased the importance of an item.

ddi:CatStatistics_1 a disco:CategoryStatistics;
  disco:frequency 13314444;
  disco:percentage 49.97;
  disco:statisticsCategory ddi:SexM;
  disco:statisticsDataFile ddi:Datafile_1.
        
ddi:CatStatistics_2 a disco:CategoryStatistics;
  disco:frequency 1336270;
  disco:statisticsCategory ddi:SexF;
  disco:statisticsDataFile ddi:Datafile_1.

Available category statistics types are frequency, percentage, and cumulativePercentage. Available summary statistics types are organized in the controlled vocabulary SummaryStatisticsType. Each summary statistics type is a skos:Concept. Particular summary statistics types are included into a disco:SummaryStatistics class with the property disco:summaryStatisticType. The particular value is modelled with rdf:value. More information on the SKOS representation of the controlled vocabulary SummaryStatisticsType can be found at the DDI-controlled-vocabularies project page. There are two possibilities to define new types of summary statistics. First, the term 'other' with a new value can be used in association with the existing vocabulary. Second, a new vocabulary can be defined. In the ISSP example below, the term 'other' is used in class issp:XYZ_17, though not included in the following tables.

There are two properties which describe details of a category or summary statistic value, computationBase and weightedBy.

computationBase expresses if the cases - which are the basis of the computation of a statistics value - are valid, invalid or the total of both. In statistics, missing data (i.e. invalid data), or missing values, occur when no data value is stored for the variable in an observation. Missing data are a common occurrence and can have a significant effect on the conclusions that can be drawn from the data. The usage of computationBase for frequency differs from the usage for the percentage statistics and the summary statistics. A distinction regarding computationBase doesn’t apply to frequency as category statistic. The following table describes the details of usage of computationBase in dependency of the respective statistics type.

Table 1: Description of Statistics of Valid/Invalid Cases

Statistics Type
computationBase
valid
invalid
total
not used
Category Statistics Type

frequency
n/a
n/a
n/a
++
percentage
++
+
++
n/a
cumulativePercentage
++
+
++
n/a
Summary Statistics Type

percentage
++
+
n/a
n/a
Any other summary statistics type
++
+
++
n/a

Legend: ++ used frequently, + rarely used, n/a not applicable

weightedBy defines the weight variable of a category or summary statistic computation respectively value. It can also be used to indicate if a weight variable is used but the related variable is not known. weightedBy may be assigned to a category statistic value or to a summary statistic value.

Table 2. Description of Statistics of Non-weighted/Weighted Variables

Statistics Value of ... Value of weightedBy
unweighted variable not used
weighted variable
Weight variable is not known.
Reference to blank node
weighted variable
Weight variable is known.
Reference to weight variable

The following example shows different categories of an ISSP data set and the values of the related summary and category statistics. Each category is defined as a skos:Concept and the used name is issp:category_X, which is the corresponding category value in the frequency table above (see Figure 23, second column).

The category issp:category_1 is the category with the code 1 (skos:notation '1'), the category label ‘Yes, have partner; live in same household’ (skos:prefLabel 'Yes, have partner; live in same household') and which is valid (disco:isValid true). Please note that the property isValid is a feature at risk, since the domain is not Disco. Maintainers of the domain ontology may introduce their own property for this or a similar purpose. issp:XYZ_1 defines the frequency (disco:frequency '15893') of the category issp:category_1 ( disco:statisticsCategory issp:category_1).

EXAMPLE: ISSP 2011 (International Social Survey Programme)
Figure 24 Example Category Statistics: Frequency Table of Variable PARTLIV (ISSP 2011)
Figure 25 Example Category Statistics: Frequency Table of Variable WRKHRS (ISSP 2011)
Figure 26 Example Summary Statistics: Descriptive Statistics of Variable WRKHRS (ISSP 2011)
@prefix issp: <http://www.issp.org/>
@prefix ddi-cv: <http://rdf-vocabulary.ddialliance.org/DDICV#>
		  
issp:Category_1
  a skos:Concept;
  skos:notation "1";
  skos:prefLabel "Yes, have partner; live in same household";
  disco:isValid true.
  
issp:Category_3
  a skos:Concept;
  skos:prefLabel "valid total";
  disco:isValid true.

issp:Category_2
  a skos:Concept;
  skos:notation "0";
  skos:prefLabel "Not available (GB))";
  disco:isValid false.

issp:Category_4
  a skos:Concept;
  skos:prefLabel "missing total";
  disco:isValid false.

issp:XYZ_1
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_1;
  disco:frequency 15893.

issp:XYZ_2
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_2;
  disco:frequency 936.
  
issp:XYZ_3
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_1;
  disco:percentage 60.6.
  disco:computationBase "total".

issp:XYZ_4
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_2;
  disco:percentage 3.6;
  disco:computationBase "total";
  disco:weightedBy issp:WeightVariable_1.
  
issp:XYZ_5
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_1;
  disco:percentage 63.7;
  disco:computationBase "validOnly".
  
issp:XYZ_6
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_1;
  disco:cumulativePercentage 63.7;
  disco:computationBase "validOnly".
  
# optional: harmonized CategoryStatistics resource if computationBase and category is the same
issp:XYZ_7
  a disco:CategoryStatistics;
  disco:statisticsCategory issp:Category_1;
  disco:percentage 63.7;
  disco:cumulativePercentage 63.7;
  disco:computationBase "validOnly".

# SummaryStatistics of variable PARTLIV
issp:XYZ_8
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:PARTLIV;
  disco:summaryStatisticType ddicv-sumstats:ValidCases;
  rdf:value "24965".

issp:XYZ_9
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:PARTLIV;
  disco:summaryStatisticType ddicv-sumstats:PercentOfValidCases;
  rdf:value "95.2".

issp:XYZ_10
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:PARTLIV;
  disco:summaryStatisticType ddicv-sumstats:InvalidCases;
  rdf:value "1251".
  
issp:XYZ_11
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:PARTLIV;
  disco:summaryStatisticType ddicv-sumstats:PercentOfInvalidCases;
  rdf:value "4.8".
  
# SummaryStatistics of variable WRKHS
issp:XYZ_12
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:ValidCases;
  rdf:value "14237".

issp:XYZ_13
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:Minimum;
  rdf:value "1".

issp:XYZ_14
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:Maximum;
  rdf:value "96".

issp:XYZ_15
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:ArithmeticMean;
  rdf:value "41.74".

issp:XYZ_16
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:StandardDeviation;
  rdf:value "14.265".

# SummaryStatistics of variable WRKHS not included in the tables
issp:XYZ_17
  a disco:SummaryStatistics;
  disco:statisticsVariable issp:WRKHRS;
  disco:summaryStatisticType ddicv-sumstats:Other;
  rdfs:label "Gini Coefficient";
  rdf:value "0.63".

Microdata Information System (MISSY)
Figure 27 Summary Statistics
Figure 28 Category Statistics
# minimum
# -------
missy:Minimum
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:Minimum;
  rdf:value "1".

# maximum
# -------
missy:Maximum
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:Maximum;
  rdf:value "4".

# arithmentic mean
# ----------------
missy:Mean
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:ArithmeticMean;
  rdf:value "2.17".

# standard deviation
# ------------------
missy:StandardDeviation
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:StandardDeviation;	
  rdf:value "0.9061".
  
# valid cases
# -----------
missy:ValidCases
  a disco:SummaryStatistics ;
  disco:statisticsVariable missy:PB100 ;
  disco:summaryStatisticType ddicv-sumstats:ValidCases;
  rdf:value "470950".

# percent of valid cases
# ----------------------
missy:PercentOfValidCases
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:PercentOfValidCases;
  rdf:value "99.1".

# invalid cases
# -------------
missy:InvalidCases
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:InvalidCases;
  rdf:value "4195".

# percent of invalid cases  
# ------------------------
missy:PercentOfInvalidCases
  a disco:SummaryStatistics ;
  disco:statisticsVariable missy:PB100 ;
  disco:summaryStatisticType ddicv-sumstats:PercentOfInvalidCases ;
  rdf:value "0.9" .
  
# total cases
# -----------
missy:TotalCases
  a disco:SummaryStatistics;
  disco:statisticsVariable missy:PB100;
  disco:summaryStatisticType ddicv-sumstats:NumberOfCases;
  rdf:value "475145".

# codes and categories
# --------------------
missy:1
  a skos:Concept ;
  skos:notation "1" ;
  skos:prefLabel "January,February,March" ;
  disco:isValid true .

missy:Missing
  a skos:Concept ;
  skos:notation "M" ;
  skos:prefLabel "Missing" ;
  disco:isValid false .

# valid cases
# -----------
missy:CS1
  a disco:CategoryStatistics ;
  disco:statisticsCategory missy:1 ;
  disco:frequency 102710 ;
  disco:percentage 21.6 ;
  disco:cumulativePercentage 21.8 ;
  disco:computationBase "valid" .

# invalid cases 
# -------------
missy:CS2
  a disco:CategoryStatistics ;
  disco:statisticsCategory missy:Missing ;
  disco:frequency 4195 ;
  disco:percentage 0.9 ;
  disco:computationBase "invalid" .

8. Variables, Variable Definitions, Representations, and Concepts

When it comes to understanding the contents of the data set, this is done using the Variable class. Variables (Variable) provide a definition of the column in a rectangular data file, and can associate it with a Concept, and a Question. Variables (Variable) are related to a Representation of some form, which may be a set of codes and categories (a "codelist") or may be one of other normal data types (dateTime, numeric, textual, etc.) Codes and Categories are represented using skos:Concept and skos:ConceptScheme. Variable definitions (RepresentedVariable) encompasse study-independent, re-usable parts of variables like occupation classification.

Figure 29 Variables, Variable Definitions, Representations, and Concepts

Variables (Variable) may be based on (basedOn) 0 or 1 variable definitions (RepresentedVariable) and variable definitions (RepresentedVariable) can be in 0 to n basedOn relationships to variables (Variable). Both variables (Variable) and variable definitions (RepresentedVariable) have Representation object properties with the class Representation as range. Variables (Variable) must have exactly 1 Representation and variable definitions (RepresentedVariable) may have 0 to n Representation connections to Representation. On the other hand, representations have 0 to n links to variable definitions (RepresentedVariable) and to variables (Variable). Variables (Variable) as well as variable definitions (RepresentedVariable) have both 1 connection to the concept which should be measured. Concepts have 0 to n relationships to variables (Variable) and variable definitions (RepresentedVariable) using the object property concept.

Disco variables are inline with statistical variables, where experiments examine the relationship between variables. In the RDF Data Cube vocabulary, variables are used as dimensions, measures, or attributes to identify and describe observations.

8.1 Variable and Variable Definition

Variables provide a definition of the column in a rectangular data file. Variable is a characteristic of a unit being observed. A variable might be the answer of a question, have an administrative source, or be derived from other variables (e.g. age group derived from age). RepresentedVariables encompasse study-independent, re-usable parts of variables like occupation classification.

Figure 30 Variables and RepresentedVariables

Variables (Variable) can be described (dcterms:description), skos:notation is used to associate names to variables and labels can be assigned to variables via the datatype property skos:prefLabel. Variable definitions (RepresentedVariable) can also be described using dcterms:description. Labels can be assigned to variable definitions (RepresentedVariable) via the datatype property skos:prefLabel. Variables (Variable) may be based on (BasedOn) 0 to 1 RepresentedVariable. BasedOn also connects variable definitions (RepresentedVariable) with 0 to n variables (Variable). Variables (Variable) and variable definitions (RepresentedVariable) are connected with exactly 1 skos:Concept via Concept. skos:Concept have this connection to 0 to n variables (Variable) and variable definitions (RepresentedVariable). Variables (Variable) are represented by 1 Representation and variable definitions (RepresentedVariable) are represented by multiple (0 - n) representations (Representation). Representations (Representation) may be linked to 0 to n variables (Variable) and their definitions. Variables (Variable) may have (Question) 0 or more questions (Question) and questions (Question) may be associated with 0 to n variables (Variable). Universe is used to link 1 Universe to 0 to n variables (Variable) and 0 to n universes (Universe) to 0 to n variable definitions (RepresentedVariable).

The following example illustrates the three variables Sex, Age and Citizenship.

   ddi:AR80A401 a disco:Variable;
     dcterms:identifier "AR80A401";
     skos:prefLabel "Sex"@en, "Sexe"@fr;
     dcterms:description "This variable indicates the person's gender."@en;
     disco:basedOn ddi:SexVD;
     disco:question ddi:QuestionGender.
         
   ddi:AR80A402 a disco:Variable;
     dcterms:identifier "AR80A402";
     dcterms:description "This variable indicates the person's age in years."@en;
     skos:prefLabel "Age"@en, "Âge"@fr.
     disco:basedOn ddi:AgeVD;
     disco:question ddi:QuestionAge.
               
   ddi:AR80A407 a disco:Variable;
     dcterms:identifier "AR80A407";
     dcterms:description "This variable indicates whether or not the person
is a naturalized citizen of Argentina."@en;
     skos:prefLabel "Citizenship"@en, "Citoyenneté"@fr;
     disco:basedOn ddi:CitizenshipVD;
     disco:question ddi:QuestionCitizenship.

The three variables refer to universe, representations and concepts in their RepresentedVariable.

 ddi:SexVD a disco:RepresentedVariable;
   disco:universe ddi:UniversePerson;
   disco:representation ddi:SexRepr;
   disco:concept ddi:IpumsC1;
   skos:prefLabel "Sex"@en, "Sexe"@fr;
   dcterms:description "Sex data element"@en.
                           
ddi:AgeVD a disco:RepresentedVariable;
   disco:universe ddi:UniversePerson;
   disco:representation ddi:AgeRepr;
   disco:concept ddi:IpumsC1;
   skos:prefLabel "Age"@en, "Sexe"@fr;
   dcterms:description "Age data element"@en.

 ddi:CitizenshipVD a disco:RepresentedVariable;
   disco:universe ddi:UniverseNonArgentines;
   disco:representation ddi:CitizenshipRepr;
   disco:concept ddi:IpumsC2;
   skos:prefLabel "Citizenship"@en;
   dcterms:description "Citizenship data element"@en.

8.2 Representation

The Representation of a variable is the combination of a value domain, datatype, and, if necessary, a unit of measure or a character set. Representation is one of a set of values to which a numerical measure or a category from a classification can be assigned (e.g. income, age, and sex: male coded as 1). Questions (ResponseDomain), variables (Variable) (Representation), and variable definitions (RepresentedVariable) (Representation) may have representations. Representation is defined as sub-class of the union of rdfs:Datatype (e.g. numeric or textual values), skos:ConceptScheme, and skos:OrderedCollection, as for example questions may have as response domain a mixture of a numeric response domain containing numeric values (rdfs:Datatype) and an unordered code response domain (skos:ConceptScheme) as well as an ordered code response domain (skos:OrderedCollection).

Figure 31 Representation

Questions (Question) (responseDomain), variables (Variable) (representation), and variable definitions ( RepresentedVariable) (representation) may have representations. Questions (Question) must have 1 to n representations (representation), variables (Variable) must have exactly 1 Representation, and variable definitions (RepresentedVariable) may have 0 to n representations (Representation). Each Representation can be in 0 to n Representation relationships with questions (Question), variables (Variable), and variable definitions (RepresentedVariable).

The following example shows the representations of the three previously introduced variables Sex, Age and Citizenship. All of them refer to the particular concepts.

  ddi:SexRepr a skos:ConceptScheme, disco:Representation;
    skos:hasTopConcept ddi:SexM, ddi:SexF.
      
ddi:AgeRepr a skos:ConceptScheme, disco:Representation;
  skos:hasTopConcept ddi:Age0, ddi:Age1, ddi:Age99.
    
ddi:CitizenshipRepr a skos:ConceptScheme, disco:Representation;
  skos:hasTopConcept ddi:CYes, ddi:CNo, ddi:CUnknown, ddi:CNIU.

8.3 Codes and Categories

DDI concepts, hierarchies of DDI concepts, code values, and category labels are represented by skos:Concepts. SKOS defines the term skos:Concept, which is a unit of knowledge created by a unique combination of characteristics. In context of statistical (meta)data, concepts are abstract summaries, general notions, knowledge of a whole set of behaviours, attitudes or characteristics which are seen as having something in common. Concepts may be associated with variables and questions. A skos:ConceptScheme, also defined within the SKOS namespace, is a set of metadata describing statistical concepts. Skos:Concept is reused to a large extent to represent DDI concepts, codes, and categories.

Figure 32 skos:Concept and skos:ConceptScheme

DDI concepts can be described using skos:definition. Furthermore, you can describe code values (skos:notation) and category labels (skos:prefLabel). Hierarchies of DDI concepts can be built using the object properties skos:broader and skos:narrower. The domains and the ranges of skos:broader and skos:narrower are skos:Concept. The cardinalities are in both directions 0 to n. Skos:Concept may be organized in 0 to n skos:ConceptSchemes by means of skos:inScheme. skos:ConceptSchemes may have multiple (0 - n) skos:Concept as parts. The top concept in a specific ConceptScheme is indicated by skos:hasTopConcept pointing to 0 to n top skos:Concept. A specific skos:Concept may be the top concept to multiple (0 - n) skos:ConceptSchemes.

ddi:SexRepr a skos:ConceptScheme, disco:Representation;
  rdfs:label "Code list for Sex (SEX) - codelist class"@en;
  rdfs:comment "This code list provides the gender."@en;
  skos:hasTopConcept ddi:SexM, ddi:SexF.
  
ddi:SexM a skos:Concept;
  skos:notation "1";
  skos:prefLabel "Male"@en, "Homme"@fr;
  skos:inScheme ddi:SexRepr.
      
ddi:SexF a skos:Concept;
  skos:notation "2";
  skos:prefLabel "Female"@en, "Femme"@fr;
  skos:inScheme ddi:SexRepr.
EXAMPLE: ISSP 2011 (International Social Survey Programme)
Figure 33 Example Category Statistics: Frequency Table of Variable PARTLIV (ISSP 2011)
@prefix issp: <http://www.issp.org/>	  
		  
issp:Category_1
  a skos:Concept;
  skos:notation "1";
  skos:prefLabel "Yes, have partner; live in same household";
  disco:isValid true.
  
issp:Category_2
  a skos:Concept;
  skos:notation "2";
  skos:prefLabel "Yes, have partner; don't live in same household";
  disco:isValid true.
  
issp:Category_3
  a skos:Concept;
  skos:notation "3";
  skos:prefLabel "No partner";
  disco:isValid true.
  
issp:Category_4
  a skos:Concept;
  disco:isValid true.

issp:Category_5
  a skos:Concept;
  skos:notation "0";
  skos:prefLabel "Not available (GB))";
  disco:isValid false.
  
issp:Category_6
  a skos:Concept;
  skos:notation "7";
  skos:prefLabel "Refused";
  disco:isValid false.
  
issp:Category_7
  a skos:Concept;
  skos:notation "9";
  skos:prefLabel "No answer";
  disco:isValid false.

issp:Category_8
  a skos:Concept;
  disco:isValid false.
Note

Please note that only code and categories are part of the turtle example.

8.4 Ordering

In DDI, variables, logical data sets, questions, and categories are typically organized themselves in a particular order. For obtaining this order, skos:OrderedCollections are used. For example, a collection of variables is represented as being of the type skos:OrderedCollection containing multiple variables (each represented as skos:Concept) in a skos:memberList.

EXAMPLE: ISSP 2011 (International Social Survey Programme)
Figure 34 Example Category Statistics: Frequency Table of Variable PARTLIV (ISSP 2011)

The following example shows an ordered collection of categories represented using abbreviated and complete syntax forms.

@prefix issp: <http://www.issp.org/>	  

issp:XYZ_1
  a disco:Variable;
  skos:notation "PARTLIV";
  skos:prefLabel "Living in steady partnership";
  disco:representation issp:OrderedCollection_1.

# abbreviated syntax:
issp:OrderedCollection_1
  rdf:type skos:OrderedCollection;
  skos:memberList (
    issp:Category_1
    issp:Category_2
    issp:Category_3
    issp:Category_4
    issp:Category_5
    issp:Category_6
    issp:Category_7
    issp:Category_8 ).
	
# complete syntax:
issp:OrderedCollection_1
  rdf:type skos:OrderedCollection;
  skos:memberList [
    rdf:first issp:Category_1; rdf:rest [
    rdf:first issp:Category_2; rdf:rest [
    rdf:first issp:Category_3; rdf:rest [
    rdf:first issp:Category_4; rdf:rest [
    rdf:first issp:Category_5; rdf:rest [
    rdf:first issp:Category_6; rdf:rest [
    rdf:first issp:Category_7; rdf:rest [
    rdf:first issp:Category_8;
    rdf:rest rdf:nil.] ] ] ] ] ] ] ].

If no order inside a collection of variables and questions is necessary, they are represented as unordered skos:ConceptSchemes. The classes Variable, LogicalDataSet, and Question are defined as sub-classes of skos:Concept.

9. Data Collection

The data collection produces the datasets in a data catalog. In some cases, where data collection is cyclic or on-going, data sets may be released as a StudyGroup, where each cycle or "wave" of the data collection activity produces one or more data sets. The data for the study are collected by an instrument. The purpose of an Instrument, i.e. an interview, a questionnaire or another entity used as a means of data collection, is in the case of a survey to record the flow of a questionnaire, its use of questions, and additional component parts. A questionnaire contains a flow of questions. A Question is designed to get information upon a subject, or sequence of subjects, from a respondent.

Figure 35 DataCollection

9.1 Instrument

The data for the study are collected by an Instrument. The purpose of an Instrument, i.e. an interview, a questionnaire or another entity used as a means of data collection, is in the case of a survey to record the flow of a questionnaire, its use of questions, and additional component parts. A questionnaire contains a flow of questions.

Instruments (Instrument) can be labeled and described using (dcterms:description) and (skos:prefLabel). Instruments (Instrument) may have (externalDocumentation) multiple (0 - n) external documentations which are of the type foaf:Documents. Foaf:Documents may be external documentations of 0 to n instruments (Instrument). collectionMode are special instruments having at least 1 (1 - n) collection mode (Question), which is a skos:Concept. A specific collection mode can be associated with 0 to n questionnaires (Questionnaire). Questionnaires (Questionnaire) must contain 1 to n questions (Question) using the object property Question. Particular questions (Question) may be contained in 0 to n questionnaires (Questionnaire).

The following example illustrates a questionnaire with three example questions. The questions are defined the next section.

ddi:Questionnaire_1 a disco:Questionnaire;
  disco:question ddi:QuestionGender;
  disco:question ddi:QuestionAge;
  disco:question ddi:QuestionCitizenship.

9.2 Question

A Question is designed to get information upon a subject, or sequence of subjects, from a respondent.

Questions (Question) have a question text (questionText), a label (skos:prefLabel), exactly 1 universe (Universe), multiple (1 - n) concepts (concept), and at least 1 response domain (responseDomain). Representations (Representation) may have 0 to n responseDomain relations to questions (Question). Particular universes (Universe) may be connected with 0 to n questions (Question). Skos:Concepts are associated with 0 to n questions (Question).

   ddi:QuestionGender a disco:Question;
     disco:questionText "2. Is the person a man or a woman? [] Man, [] Woman"@en.
     
   ddi:questionAge a disco:Question;
     disco:QuestionText "3. What is his or her age? _ _ Mark the age in
completed years at the date of the census for those younger than
one year old mark 00. For those younger than 10 years old, mark 01,
02, 03, etc. For those older than 99 years old, mark 99."@en.
       
   ddi:questionCitizenship a disco:Question;
     disco:QuestionText "6. [Immigration status] Only for persons who have
usual residence in Argentina and were born in another country.
[Questions 6A and 6B asked only of persons born outside Argentina
and who currently reside in Argentina.] B. Are you a naturalized
citizen of Argentina? [] Yes [] No [] Unanswered"@en.

10. Use of Other Vocabularies

Widely accepted and adopted vocabularies are reused to a large extend. Many features of DDI can be addressed by classes and properties of other vocabularies, such as: describing metadata for citation purposes using the DCMI Metadata Terms (DCMI) [DCMI], describing catalogues of datasets using the Data Catalog Vocabulary (DCAT) [DCAT], describing aggregate data like multi-dimensional tables using the RDF Data Cube Vocabulary [RDF Data Cube Vocabulary], describing formal statistical classifications using the SKOS Extension for Statistics (XKOS) [XKOS], describing arbitrary (real, hypothesized, virtual, fictional) objects, processes and their attributes [SIO], and delineating code lists, category schemes, mappings between them, and concepts like topics using the Simple Knowledge Organization System (SKOS) [SKOS]. Furthermore, the external vocabularies Friend of a Friend (FOAF) [FOAF], the Organization Ontology (ORG) [ORG], the Asset Description Metadata Schema (ADMS) [ADMS], and the PROV Ontology (PROV-O) [PROV-O] are used. Whenever terms from other vocabularies are used within the Disco context, these terms are not re-defined but only applied for the purposes of disco.

It is distinguished between required, recommended and optional vocabularies that are reused. Required vocabularies contain classes and properties that are required in order to represent particular aspects of Disco completely. Recommended vocabularies hold classes and properties that are recommended to be used for representing particular aspects of Disco. Finally, optional vocabularies contain classes and properties that may support the modelling of particular aspects of Disco. This strongly depends on in which extent and for which purpose data is represented in Disco. Terms of optional vocabularies are not necessarily required for representing DDI metadata in Disco.

Required vocabularies are:

Recommended vocabularies are: Optional vocabularies are:

10.1 DCMI Metadata Terms (DCMI)

DCMI is reused in order to describe general metadata of Disco constructs such as a study abstract (dcterms:abstract), a study or dataset title (dcterms:title), a human readable description of a Disco construct (dcterms:description), provenance information for a data file (dcterms:provenance), or the date (or date range) at which a study will become available (dcterms:available).

10.2 Simple Knowledge Organization System (SKOS)

skos:Concept is reused to a large extent to represent DDI concepts, codes, and categories. SKOS defines the term skos:Concept, which is a unit of knowledge created by a unique combination of characteristics. In context of statistical (meta)data, concepts are abstract summaries, general notions, knowledge of a whole set of behaviours, attitudes or characteristics which are seen as having something in common. Skos:Concepts may be associated with variables, variable definitions, and questions and are reused to a large extent to represent DDI concepts (skos:prefLabel), codes (skos:notation), and category labels (skos:prefLabel). Skos:Concepts may be organized in skos:ConceptSchemes (skos:inScheme), sets of metadata describing statistical concepts. Hierarchies of DDI concepts can be built using the object properties skos:broader and skos:narrower. Topical coverage can be expressed using dcterms:subject. Disco foresees the use of skos:Concept for the description of topical coverage. Spatial, temporal, and topical coverage are directly attached to studies, logical datasets, and datafiles. Universes and AnalysisUnits are also skos:Concepts. Therefore the properties defined for skos:Concept can be reused. KindOfData, pointing to a skos:Concept , describes, with a string or a term from a controlled vocabulary, the kind of data documented in the logical product(s) of a Study. Using dcterms:format, DataFiles formats can be defined.

10.2.1 Uses of skos:Concept

In this sub-section, we describe all possible uses of the class skos:Concept.
  • Code values: Code values are represented using the datatype property skos:notation with skos:Concept as domain.
  • Category labels: Use skos:prefLabel and the domain class skos:Concept to describe category values
  • DDI concepts: DDI concepts are described by the property skos:definition pointing from skos:Concept classes.
  • Hierarchies of DDI concepts: Hierarchies of DDI concepts can be built using the object properties skos:broader and skos:narrower. The domains and the ranges of skos:broader and skos:narrower are skos:Concept.
  • Organization in skos:ConceptSchemes: Skos:Concepts may be organized in skos:ConceptSchemes by means of skos:inScheme. The top concept in a specific ConceptScheme is indicated by skos:hasTopConcept pointing to top skos:Concept.
  • Topical coverage: Topical coverage can be expressed using dcterms:subject. DDI-RDF foresees the use of skos:Concept for the description of topical coverage. Spatial, temporal, and topical coverage are directly attached to studies, logical datasets, and datafiles.
  • Category linked to CategoryStatistics: CategoryStatistics like frequencies and percentages are associated to the respectve Category using the object property statisticsCategory. skos:Concept represents categories.
  • Concepts of questions: Questions (Question) are associated with concepts via the object property concept.
  • Universe: Each universe is also a skos:Concept. Therefore the properties defined for skos:Concept can be reused for universes.
  • Collection Mode: Questionnaires (Questionnaire) may have multiple collection modes which are represented by skos:Concept.
  • Concepts of variable definitions: Variable definitions are associated with concepts via the object property concept.
  • Concepts of variables: Variables (Variable) are linked to concepts via the object property concept.
  • Kind of data: KindOfData describes, with a string or a term from a controlled vocabulary, the kind of data documented in the logical product(s) of a Study. Examples include survey data, census/enumeration data, administrative data, measurement data, assessment data, demographic data, voting data, etc. The range of kindOfData is skos:Concept
  • Format of data files: Using the object property dcterms:format, data files (DataFile) formats can be defined. Data files (DataFiles) must have exactly 1 dcterms:format relationship to an instance of the class dcterms:MediaTypeOrExtend which is a sub-class of skos:Concept.
  • AnalysisUnit: Each analysis unit is also a skos:Concept. Therefore the properties defined for skos:Concept can be reused for analysis units.

10.3 Data Catalog Vocabulary (DCAT)

DCAT is a W3C standard for describing catalogs of datasets. DCAT makes few assumptions about the kind of datasets being described, and focuses on general metadata about the datasets (mostly using Dublin Core), and on different ways of distributing and accessing the dataset, including availability of the dataset in multiple formats. Combining terms from both DCAT and Disco can be useful for a number of reasons:

The LogicalDataSet is an extension of the dcat:DataSet. Physical, distributed files are represented by the DataFile, which is itself an extension of dcat:Distribution.

ddi:DataCatalog_1
a dcat:Catalog;
dcat:record ddi:EuropeanStudy;
dcat:dataset ddi:EuropeanDataset;
          
ddi:EuropeanStudy
a dcat:CatalogRecord;
a disco:Study;
foaf:primaryTopic ddi:EuropeanDataset;
disco:product ddi:EuropeanDataset.
          
ddi:EuropeanDataset
a dcat:Dataset;
a disco:LogicalDataSet;
dcat:theme ddi:topics/WellBeing;
dcat:theme ddi:topics/PoliticalAttitudes;
dcat:keyword "Europe"@en;
dcat:keyword "Politics"@en.

10.4 Friend of a Friend (FOAF) and Organization Ontology (ORG)

Within the context of Disco, FOAF as well as ORG are reused. Creators (dcterms:creator), contributors (dcterms:contributor), and publishers (dcterms:publisher) of Studies and StudyGroups are foaf:Agents which are either foaf:Persons or org:Organizations whose members are foaf:Persons. Studies and StudyGroups may be funded by (disco:fundedBy) foaf:Agents. The object property disco:fundedBy is defined as sub-property of dcterms:contributor.

10.5 Asset Description Metadata Schema (ADMS)

Especially persons and organizations may hold one or more persistent identifiers of particular schemes and agencies (e.g. ORCID, FundRef) that are not considered by the specific IDs of Disco. In order to include those identifiers and for distinguishing between multiple identifiers for the same class, ADMS is utilized. As a profile of DCAT, ADMS aims to describe semantic assets, i.e. reusable metadata and reference data. The class adms:Identifier can be added to a rdfs:Resource by using the property adms:identifier. That identifier class can contain properties that define the particular identifier itself, but also its scheme, version and managing agency. However, although utilized primarily for describing identifiers of persons and organizations, it is allowed to attach an adms:Identifier class to all classes in Disco.

10.6 PROV Ontology (PROV-O)

In order to represent detailed provenance information of Web data and metadata, classes and properties of PROV-O can be used. Thus, it can be used as a natural vocabulary to attach provenance information to Disco metadata. Terms of PROV-O are organized among three main classes: prov:Entity, prov:Activity and prov:Agent. While classes of Disco can be represented either as entities or agents, particular processes for, e.g. creating, maintaining and accessing data can be modeled as activities. Properties like prov:wasGeneratedBy, prov:hadPrimarySource, prov:wasInvalidatedBy, or prov:wasDerivedFrom describe the relationship between classes for the generation of data in more detail. In order to link from a disco:Study to its original DDI XML file, the property prov:wasDerivedFrom can be used. Moreover, PROV-O allows for representing versioning information by e.g., using the terms prov:Revision, prov:hadGeneration and prov:hadUsage.

10.7 RDF Data Cube Vocabulary (QB)

The RDF Data Cube Vocabulary is a W3C standard for representing data cubes, that is, multidimensional aggregate data. A qb:DataSet represents aggregate data such as multi-dimensional tables. Aggregate data is derived from microdata by statistics on groups, or aggregates such as counts, means, or frequencies. Data cubes are often generated by tabulating or aggregating unit-record datasets. For example, if an observation in a census data cube indicates the population of a certain age group in a certain region is 12345, then this fact was obtained by aggregating that number of individual records from a unit-record dataset. Disco contains a property “aggregation” that indicates that a Cube dataset was derived by tabulating a unit-record dataset. Data Cube provides for the description of the structure of such cubes, but also for the representation of the cube data itself, that is, the observations that make up the cube dataset [Semantic Statistics]. This is not the case for Disco, which only describes the structure of a dataset, but is not concerned with representing the actual data in it. The actual data are assumed to sit in a data file (e.g. a CSV file, or in a proprietary statistical package file format) that is not represented in RDF.

10.7.1 Examples

Simple case: provenance of aggregated data / relationship from aggregated data to microdata
Simple case
@prefix prov: <http://www.w3.org/ns/prov#> .
							
ddi:AggregatedDataSet
    a prov:Entity;
    prov:wasDerivedFrom ddi:MicrodataDataSet.

ddi:MicrodataDataSet a prov:Entity .
Complex case: detailed description of microdata variables resulting dimensions in aggregated data and aggregation method
Complex case
@prefix prov: <http://www.w3.org/ns/prov#> .
							
ddi:AggregatedDataSet
    a prov:Entity;
    prov:wasDerivedFrom ddi:MicrodataDataSet;
    prov:wasGeneratedBy ddi:AggregationActivity;
    prov:qualifiedDerivation [
        a prov:Derivation;
        prov:entity ddi:MicrodataDataSet;
        prov:hadActivity ddi:AggregationActivity ].

ddi:AggregationActivity
    a prov:Activity .

ddi:MicrodataDataSet
    a prov:Entity;
Note
  • prov:Activity
    • reference to aggregation method described by CV
    • description of variables in the microdata data set / 0..n independent variables / 1..n dependent variables
    • description of the dimension in data cube
    • 1 activity per dimension
    • see more information in DDI-L field level documentation DDI 3.2 Aggregation

10.8 SKOS Extension for Statistics (XKOS)

The use of formal statistical classifications is very common in research datasets - these are treated in Disco as SKOS concepts, but in some cases those working with formal statistical classifications may desire more expressive capability than SKOS provides. To support such users, the DDI Alliance also develops XKOS, a vocabulary which extends SKOS to allow for a more complete description of such classifications [eXtended Knowledge Organization System]. While the use of XKOS is not required by this vocabulary, the two are designed to work in complementary fashion. SKOS properties may be substituted by additional XKOS properties.

Which Datasets Have A Specific Statistical Classification and What Are Its Semantic Relations?

XKOS extends SKOS with two main objectives: the first one is to allow the description of statistical classifications, the second one is to introduce refinements of the semantic properties defined in SKOS. The semantic properties extend the possible relations that can be applied between pairs of skos:Concepts. SKOS allows the following relations: skos:broader than, skos:narrower than, and skos:related to. The first two are hierarchical relations, one in each direction. In Disco, these SKOS properties may be substituted by additional XKOS properties like xkos:generalizes, xkos:hasPart, xkos:caused, xkos:previous, and xkos:next.

One question, typically asked by social science researchers, could be to query all the datasets (disco:LogicalDataSet) which have a specific statistical classification (skos:ConceptScheme) like ISCO (International Standard Classification of Occupations) or ANZSIC (Australian and New Zealand Industry Classification). It is also possible to query on the semantic relationships which are defined for statistical classifications using XKOS properties. By means of these properties not only hierarchical relations can be queries but also for example part of relationships (xkos:hasPart), more general (xkos:generalizes) and more specific (xkos:specializes) concepts, and positions of concepts in lists (xkos:previous, xkos:next).

10.9 Semanticscience Integrated Ontology (SIO)

The Semanticscience Integrated Ontology (SIO) provides a simple, integrated ontology of types and relations for rich description of objects, processes and their attributes. A sio:SIO_000367 (Variable) represents a value that may change within the scope of a given or set of operations. For instance, in the context of a mathematics or statistics, a sio Variable is an information content entity that can be used to indicate the independent, dependent, or control variables of a study or experiment. Here, the similarity between sio Variable and disco:Variable is that, they are both associated to a concept e.g., Sex, Age and Citizenship.

11. DDI-XML Bidirectional Mappings

The main intention of Disco is to provide a RDF representation of DDI resources for discovery purposes in the Linked Data web. Nevertheless, bidirectional mappings between disco and DDI Lifecycle (DDI-L) are provided. In this section, bidirectional mappings between Disco and DDI Lifecycle (DDI-L) is provided. It allows an easy adoption of the DDI Discovery Vocabulary for existing DDI metadata. XSLTs for converting any XML output of DDI Codebook (DDI-C) and DDI-L are available at the DDI-RDF-tools project page.

Official Mapping Document

There is also an official document containing all bidirectional mappings between Disco and DDI-L: official mapping document These mapping tables will be transformed to the official specification in form of a turtle file and in form of html tables in this html specification.

Bidirectional Mappings between Disco and DDI-L Mappings between Disco and other Versions of DDI-XML

In order to avoid inconsistencies (as mapping tables may changes over time), we only offer mappings between Disco and the concrete version DDI 3.1 of DDI-L. There are various mapping documents between DDI 3.1 and other DDI versions (like DDI 3.2 and DDI 2.1) on the DDI Alliance website.

Mappings between Disco and DDI 4

DDI 4 will be the next model-driven specification of DDI including mappings to multiple representations such as RDF, XML, relational databases, and Java. DDI 4 should have a clear mapping from DDI-XML 3.2. We assume that all items used in Disco will have a clear mapping to DDI-XML 3.2, and these items in DDI-XML 3.2 will have a clear mapping to items in the DDI 4 model (therefore to a representation in OWL/RDF as well). If the latter should not be possible, then a mapping of items in DDI-XML 3.2 to DDI 4 XML and DDI 4 RDF should be possible.

Turtle File Containing Mappings in RDF

The mappings are defined within a separate turtle file

Mapping Tables

11.1 Representation of Mappings in RDF

Figure 36 Representation of Mappings in RDF
Mapping Examples
skos:notation a rdfs:Class, owl:Class ;
disco:mapping [
    a disco:Mapping ;
    disco:ddi-L-XPath "//l:Variable/l:VariableName" ;
    disco:ddi-L-Documentation "http://www.ddialliance.org/Specification/DDI-Lifecycle/3.1/XMLSchema/FieldLevelDocumentation/logicalproduct_xsd/elements/Variable.html" 
    disco:context "skos:notation represents variable label" ;
    disco:context "SELECT ?notation WHERE { ?notation rdfs:domain ?variable. ?variable a disco:Variable. }" ]

11.2 Classes

11.2.1 Disco

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 disco:AnalysisUnit r:AnalysisUnit
#2 disco:RepresentedVariable
#3 disco:DataFile
#4 disco:DescriptiveStatistics
#5 disco:SummaryStatistics
#6 disco:CategoryStatistics p:CategoryStatistics
#7 disco:Instrument d:Instrument
#8 disco:LogicalDataSet
#9 disco:Question d:QuestionItem | d:MultipleQuestionItem
#10 disco:responseDomain
#11 disco:Questionnaire d:Instrument The instument of the study
#12 disco:Study s:StudyUnit
#13 disco:StudyGroup
#14 disco:Variable //l:Variable

11.2.2 External

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 skos:ConceptScheme //l:Variable/l:CodeScheme Variables can have a coded representaion

11.3 Object Properties

11.3.1 Disco

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 disco:analysisUnit
#2 disco:basedOn
#3 disco:collectionMode
#4 disco:variable
#5 disco:concept //l:Vaiable/l:ConceptReference Varialbe has a concept
#6 disco:concept //d:QuestionItem/r:ConceptReference Question is defined by concept
#7 "
#8 disco:aggregation
#9 disco:dataFile
#10 disco:ddifile
#11 disco:externalDocumentation
#12 disco:fundedBy
#13 disco:inGroup
#14 disco:inputVariable
#15 disco:instrument //d:DataCollection/[d:QuestionItem d:MultipleQuestionItem] The instrument of the study questionaire
#16 disco:kindOfData
#17 disco:product
#18 disco:question //l:Variable/l:QuestionReference Variable can have a question
#19 disco:question //[d:QuestionItem d:MultipleQuestionItem] Questions in a questionaire
#20 disco:representation //l:Variable/l:Representation/l:CodeRepresentation/[r:CodeSchemeReference l:NumericRepresentation l:TextRepresentation l:DateTimeRepresentation] Variables can have a representation
#21 disco:statisticsCategory
#22 disco:statisticsDataFile
#23 disco:statisticsVariable
#24 disco:weightedBy
#25 disco:universe disco:universe Variable can have a concept

11.3.2 External

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 dcterms:identifier //l:Variable/l:VariableName dcterms:identifier represents variable label
#2 skos:prefLabel //l:Variable/r:Label skos:prefLabel represents the label of the variable
#3 skos:prefLabel //d:QuestionItem/d:QuestionItemName Name of question

11.4 Data Properties

11.4.1 Disco

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 skos:notation //l:Variable/l:VariableName skos:notation represents variable label DDI-L Documentation
#2 disco:frequency p:CaseQuantity
#3 disco:isPublic
#4 disco:isValid
#5 disco:questionText d:QuestionText
#6 disco:percentage
#7 disco:computationBase
#8 disco:cumulativePercentage
#9 disco:purpose s:Purpose
#10 disco:subtitle r:SubTitle
#11 disco:standardDeviation
#12 disco:numberOfCases
#13 disco:maximum
#14 disco:mean
#15 disco:median
#16 disco:minimum
#17 disco:mode
#18 disco:startDate

11.4.2 External

#
property
domain class
range class
DDI-L
description
DDI-L Documentation
#1 skos:notation //l:Variable/l:VariableName skos:notation represents variable label DDI-L Documentation
#2 skos:notation skos:notation represents code

11.5 Overview of the Mapping from DDI-C and DDI-L to DDI-RDF

11.5.1 Studies and StudyGroups

#
property
domain class
range class
DDI-C
DDI-L
1 universe union of Study and StudyGroup Universe X X
2 dcterms:subject union of Study and StudyGroup skos:Concept X
3 dcterms:temporal union of Study and StudyGroup dcterms:PeriodOfTime
4 dcterms:spatial union of Study and StudyGroup dcterms:Location
5 kindOfData union of Study and StudyGroup skos:Concept X
6 analysisUnit union of Study and StudyGroup AnalysisUnit
7 dcterms:abstract union of Study and StudyGroup rdf:langString X X
8 dcterms:alternative union of Study and StudyGroup rdf:langString X X
9 dcterms:available union of Study and StudyGroup xsd:dateTime X
10 dcterms:title union of Study and StudyGroup rdf:langString X X
11 purpose union of Study and StudyGroup rdf:langString X
12 subtitle union of Study and StudyGroup rdf:langString X X
13 ddiFile union of Study and StudyGroup foaf:Document
14 fundedBy union of Study and StudyGroup foaf:Agent
15 dcterms:creator union of Study and StudyGroup foaf:Agent X
16 dcterms:contributor union of Study and StudyGroup foaf:Agent
17 dcterms:publisher union of Study and StudyGroup foaf:Agent - X
18 instrument Study Instrument X
19 inGroup Study StudyGroup X
20 dataFile Study DataFile X
21 variable Study Variable X X
22 product Study LogicalDataSet X
23 owl:versionInfo Study
24 skos:definition Universe rdf:langString X

11.5.2 General Metadata

#
property
domain class
range class
DDI-C
DDI-L
1 adms:identifier disco:Study adms:Identifier X
2 adms:identifier disco:StudyGroup adms:Identifier
3 adms:identifier disco:AnalysisUnit adms:Identifier
4 adms:identifier disco:Universe adms:Identifier
5 adms:identifier disco:LogicalDataSet adms:Identifier
6 adms:identifier disco:DataFile adms:Identifier X
7 adms:identifier disco:DescriptiveStatistics adms:Identifier
8 adms:identifier disco:SummaryStatistics adms:Identifier
9 adms:identifier disco:CategoryStatistics adms:Identifier
10 adms:identifier disco:Variable adms:Identifier X
11 adms:identifier disco:RepresentedVariable adms:Identifier
12 adms:identifier disco:Question adms:Identifier
13 adms:identifier disco:Instrument adms:Identifier
14 adms:identifier disco:Questionnaire adms:Identifier
15 skos_prefLabel rdfs:Resource rdf:langString
16 dcterms:relation rdfs:Resource foaf:Document
17 dcterms:description dcterms:RightsStatement rdf:langString
18 skos:prefLabel dcterms:RightsStatement rdf:langString
19 rdfs:seeAlso dcterms:RightsStatement foaf:Document
20 skos:prefLabel dcterms:PeriodOfTime rdf:langString
21 startDate dcterms:PeriodOfTime xsd:date
22 endDate dcterms:PeriodOfTime xsd:Date
23 skos:prefLabel dcterms:MediaTypeOrExtent rdf:langString
24 org:memberOf foaf:Person org:Organization

11.5.3 Data Sets, Data Files, and Descriptive Statistics

#
property
domain class
range class
DDI-C
DDI-L
1 instrument LogicalDataSet Instrument
2 dataFile LogicalDataSet DataFile
3 aggregation LogicalDataSet qb:DataSet
4 variable LogicalDataSet Variable
5 universe LogicalDataSet Universe X
6 dcterms:title LogicalDataSet rdf:langString X
7 isPublic LogicalDataSet xsd:boolean
8 dcterms:accessRights LogicalDataSet dcterms:RightsStatement X
9 dcterms:license LogicalDataSet dcterms:LicenseDocument
10 inputVariable qb:DataSet Variable
11 caseQuantity DataFile xsd:nonNegativeInteger X
12 dcterms:description DataFile rdf:langstring
13 owl:versioninfo DataFile string X
14 dcterms:temporal DataFile dcterms:PeriodOfTime
15 dcterms:spatial DataFile dcterms:Location X
16 dcterms:provenance DataFile dcterms:ProvenanceStatement
17 dcterms:subject DataFile skos:Concept
18 dcterms:format DataFile dcterms:MediaTypeOrExtend
19 statisticsDataFile DescriptiveStatistics DataFile
20 statisticsVariable SummaryStatistics Variable
21 invalidcases SummaryStatistics xsd:nonNegativeInteger
22 maximum SummaryStatistics xsd:decimal
23 mean SummaryStatistics xsd:decimal
24 median SummaryStatistics xsd:decimal
25 minimum SummaryStatistics xsd:decimal
26 mode SummaryStatistics xsd:decimal
27 standardDeviation SummaryStatistics xsd:decimal
28 validCases SummaryStatistics xsd:nonNegativeInteger
29 weightedInvalidCases SummaryStatistics xsd:nonNegativeInteger
30 weightedMean SummaryStatistics xsd:decimal
31 weightedMedian SummaryStatistics xsd:decimal
32 weightedMode SummaryStatistics xsd:decimal
33 weightedValidCases SummaryStatistics xsd:nonNegativeInteger
34 statisticsCategory CategoryStatistics skos:Concept
35 cumulativePercentage CategoryStatistics xsd:decimal
36 frequency CategoryStatistics xsd:nonNegativeInteger
37 percentage CategoryStatistics xsd:decimal
38 weightedCumulativePercentage CategoryStatistics xsd:decimal
39 weightedFrequency CategoryStatistics xsd:nonNegativeInteger
40 weightedPercentage CategoryStatistics xsd:decimal

11.5.4 Variables, Variable Definitions, Representations, and Concepts

#
property
domain class
range class
DDI-C
DDI-L
1 skos:inScheme skos:Concept skos:ConceptScheme
2 skos:hasTopConcept skos:ConceptScheme skos:Concept
3 skos:broader skos:Concept skos:Concept X
4 skos:narrower skos:Concept skos:Concept
5 skos:definition skos:Concept rdf:langString
6 skos:notation skos:Concept rdfs:Literal X
7 skos:prefLabel skos:Concept rdf:LangString
8 question Variable Question X
9 universe Variable Universe X X
10 analysisUnit Variable AnalysisUnit
11 concept Variable skos:Concept X
12 representation Variable Representation
13 basedOn Variable RepresentedVariable
14 dcterms:description Variable rdf:langString X
15 skos:notation Variable rdfs:Literal X
16 skos:prefLabel Variable rdf:langString X
17 concept RepresentedVariable skos:Concept
18 universe RepresentedVariable Universe
19 representation RepresentedVariable Representation
20 dcterms:description RepresentedVariable rdf:langString
21 skos:prefLabel RepresentedVariable rdf:langString

11.5.5 Data Collection

#
property
domain class
range class
DDI-C
DDI-L
1 universe Question Universe X X
2 concept Question skos:Concept X
3 responseDomain Question Representation
4 questionText Question rdf:langString X
5 skos:prefLabel Question rdf:langString X
6 question Questionnaire Question
7 collectionMode Questionnaire skos:Concept
8 externalDocumentation Instrument foaf:Document
9 dcterms:description Instrument rdf:langString X
10 skos:prefLabel Instrument rdf:langString X

11.6 Mapping from DDI-C to DDI-RDF

11.6.1 Studies and StudyGroups

#
property
domain class
range class
mapping
1 universe union of Study and StudyGroup Universe /codeBook/stdyDscr/stdyInfo/sumDscr/universe
2 dcterms:subject union of Study and StudyGroup skos:Concept
3 dcterms:temporal union of Study and StudyGroup dcterms:PeriodOfTime
4 dcterms:spatial union of Study and StudyGroup dcterms:Location
5 kindOfData union of Study and StudyGroup skos:Concept
6 analysisUnit union of Study and StudyGroup AnalysisUnit
7 dcterms:abstract union of Study and StudyGroup rdf:langString /codeBook/stdyDscr/stdyInfo/abstract
8 dcterms:alternative union of Study and StudyGroup rdf:langString /codeBook/stdyDscr/citation/altTitl
9 dcterms:available union of Study and StudyGroup xsd:dateTime
10 dcterms:title union of Study and StudyGroup rdf:langString /codeBook/stdyDscr/citation/titl
11 purpose union of Study and StudyGroup rdf:langString
12 subtitle union of Study and StudyGroup rdf:langString /codeBook/stdyDscr/citation/subTitl
13 ddiFile union of Study and StudyGroup foaf:Document
14 fundedBy union of Study and StudyGroup foaf:Agent
15 dcterms:creator union of Study and StudyGroup foaf:Agent
16 dcterms:contributor union of Study and StudyGroup foaf:Agent
17 dcterms:publisher union of Study and StudyGroup foaf:Agent
18 instrument Study Instrument
19 inGroup Study StudyGroup
20 dataFile Study DataFile
21 variable Study Variable /codeBook/dataDscr/var/@id
22 product Study LogicalDataSet
23 owl:versionInfo Study
24 skos:definition Universe rdf:langString

notes

  • (1): -

11.6.2 General Metadata

#
property
domain class
range class
mapping
1 adms:identifier disco:Study adms:Identifier
2 adms:identifier disco:StudyGroup adms:Identifier
3 adms:identifier disco:AnalysisUnit adms:Identifier
4 adms:identifier disco:Universe adms:Identifier
5 adms:identifier disco:LogicalDataSet adms:Identifier
6 adms:identifier disco:DataFile adms:Identifier
7 adms:identifier disco:DescriptiveStatistics adms:Identifier
8 adms:identifier disco:SummaryStatistics adms:Identifier
9 adms:identifier disco:CategoryStatistics adms:Identifier
10 adms:identifier disco:Variable adms:Identifier
11 adms:identifier disco:RepresentedVariable adms:Identifier
12 adms:identifier disco:Question adms:Identifier
13 adms:identifier disco:Instrument adms:Identifier
14 adms:identifier disco:Questionnaire adms:Identifier
15 skos_prefLabel rdfs:Resource rdf:langString
16 dcterms:relation rdfs:Resource foaf:Document
17 dcterms:description dcterms:RightsStatement rdf:langString
18 skos:prefLabel dcterms:RightsStatement rdf:langString
19 rdfs:seeAlso dcterms:RightsStatement foaf:Document
20 skos:prefLabel dcterms:PeriodOfTime rdf:langString
21 startDate dcterms:PeriodOfTime xsd:date
22 endDate dcterms:PeriodOfTime xsd:Date
23 skos:prefLabel dcterms:MediaTypeOrExtent rdf:langString
24 org:memberOf foaf:Person org:Organization

notes

  • (1): -

11.6.3 Data Sets, Data Files, and Descriptive Statistics

#
property
domain class
range class
mapping
1 instrument LogicalDataSet Instrument
2 dataFile LogicalDataSet DataFile
3 aggregation LogicalDataSet qb:DataSet
4 variable LogicalDataSet Variable
5 universe LogicalDataSet Universe /codeBook/stdyDscr/stdyInfo/sumDscr/universe
6 dcterms:title LogicalDataSet rdf:langString
7 isPublic LogicalDataSet xsd:boolean
8 dcterms:accessRights LogicalDataSet dcterms:RightsStatement
9 dcterms:license LogicalDataSet dcterms:LicenseDocument
10 inputVariable qb:DataSet Variable
11 caseQuantity DataFile xsd:nonNegativeInteger
12 dcterms:description DataFile rdf:langstring
13 owl:versioninfo DataFile string
14 dcterms:temporal DataFile dcterms:PeriodOfTime
15 dcterms:spatial DataFile dcterms:Location
16 dcterms:provenance DataFile dcterms:ProvenanceStatement
17 dcterms:subject DataFile skos:Concept
18 dcterms:format DataFile dcterms:MediaTypeOrExtend
19 statisticsDataFile DescriptiveStatistics DataFile
20 statisticsVariable SummaryStatistics Variable
21 invalidcases SummaryStatistics xsd:nonNegativeInteger
22 maximum SummaryStatistics xsd:decimal
23 mean SummaryStatistics xsd:decimal
24 median SummaryStatistics xsd:decimal
25 minimum SummaryStatistics xsd:decimal
26 mode SummaryStatistics xsd:decimal
27 standardDeviation SummaryStatistics xsd:decimal
28 validCases SummaryStatistics xsd:nonNegativeInteger
29 weightedInvalidCases SummaryStatistics xsd:nonNegativeInteger
30 weightedMean SummaryStatistics xsd:decimal
31 weightedMedian SummaryStatistics xsd:decimal
32 weightedMode SummaryStatistics xsd:decimal
33 weightedValidCases SummaryStatistics xsd:nonNegativeInteger
34 statisticsCategory CategoryStatistics skos:Concept
35 cumulativePercentage CategoryStatistics xsd:decimal
36 frequency CategoryStatistics xsd:nonNegativeInteger
37 percentage CategoryStatistics xsd:decimal
38 weightedCumulativePercentage CategoryStatistics xsd:decimal
39 weightedFrequency CategoryStatistics xsd:nonNegativeInteger
40 weightedPercentage CategoryStatistics xsd:decimal

notes

  • (1): -

11.6.4 Variables, Variable Definitions, Representations, and Concepts

#
property
domain class
range class
mapping
1 skos:inScheme skos:Concept skos:ConceptScheme
2 skos:hasTopConcept skos:ConceptScheme skos:Concept
3 skos:broader skos:Concept skos:Concept
4 skos:narrower skos:Concept skos:Concept
5 skos:definition skos:Concept rdf:langString
6 skos:notation skos:Concept rdfs:Literal
7 skos:prefLabel skos:Concept rdf:LangString
8 question Variable Question
9 universe Variable Universe /codeBook/stdyDscr/stdyInfo/sumDscr/universe
10 analysisUnit Variable AnalysisUnit
11 concept Variable skos:Concept
12 representation Variable Representation
13 basedOn Variable RepresentedVariable
14 dcterms:description Variable rdf:langString
15 skos:notation Variable rdfs:Literal
16 skos:prefLabel Variable rdf:langString
17 concept RepresentedVariable skos:Concept
18 universe RepresentedVariable Universe
19 representation RepresentedVariable Representation
20 dcterms:description RepresentedVariable rdf:langString
21 skos:prefLabel RepresentedVariable rdf:langString

notes

  • (1): -

11.6.5 Data Collection

#
property
domain class
range class
mapping
1 universe Question Universe /codeBook/stdyDscr/stdyInfo/sumDscr/universe
2 concept Question skos:Concept
3 responseDomain Question Representation
4 questionText Question rdf:langString
5 skos:prefLabel Question rdf:langString
6 question Questionnaire Question
7 collectionMode Questionnaire skos:Concept
8 externalDocumentation Instrument foaf:Document
9 dcterms:description Instrument rdf:langString
10 skos:prefLabel Instrument rdf:langString

notes

  • (1): -

11.7 Mapping from DDI-L to DDI-RDF

11.7.1 Studies and StudyGroups

#
property
domain class
range class
mapping
1 universe union of Study and StudyGroup Universe /ddi:DDIInstance/s:StudyUnit/r:UniverseReference/r:ID
2 dcterms:subject union of Study and StudyGroup skos:Concept /ddi:DDIInstance/s:StudyUnit/r:TopicalCoverage/r:Subject
3 dcterms:temporal union of Study and StudyGroup dcterms:PeriodOfTime
4 dcterms:spatial union of Study and StudyGroup dcterms:Location
5 kindOfData union of Study and StudyGroup skos:Concept /ddi:DDIInstance/s:StudyUnit/r:KindOfData
6 analysisUnit union of Study and StudyGroup AnalysisUnit /ddi:DDIInstance/s:StudyUnit/r:AnalysisUnit
7 dcterms:abstract union of Study and StudyGroup rdf:langString /ddi:DDIInstance/s:StudyUnit/s:Abstract/r:Content
8 dcterms:alternative union of Study and StudyGroup rdf:langString /ddi:DDIInstance/s:StudyUnit/r:Citation/r:AlternateTitle
9 dcterms:available union of Study and StudyGroup xsd:dateTime /ddi:DDIInstance/s:StudyUnit/r:Embargo/r:Date/r:SimpleDate
10 dcterms:title union of Study and StudyGroup rdf:langString /ddi:DDIInstance/s:StudyUnit/r:Citation/r:Title
11 purpose union of Study and StudyGroup rdf:langString /ddi:DDIInstance/s:StudyUnit/s:Purpose/r:Content
12 subtitle union of Study and StudyGroup rdf:langString /ddi:DDIInstance/s:StudyUnit/r:Citation/r:SubTitle
13 ddiFile union of Study and StudyGroup foaf:Document
14 fundedBy union of Study and StudyGroup foaf:Agent /ddi:DDIInstance/s:StudyUnit/r:FundingInformation
15 dcterms:creator union of Study and StudyGroup foaf:Agent /ddi:DDIInstance/s:StudyUnit/r:Citation/r:Creator
16 dcterms:contributor union of Study and StudyGroup foaf:Agent /ddi:DDIInstance/s:StudyUnit/r:Citation/r:Contributor
17 dcterms:publisher union of Study and StudyGroup foaf:Agent /ddi:DDIInstance/s:StudyUnit/r:Citation/r:Publisher
18 instrument Study Instrument /ddi:DDIInstace/s:StudyUnit/d:DataCollection/@id
19 inGroup Study StudyGroup //s:StudyUnit/ancestor::g:Group[1]/@id
20 dataFile Study DataFile //s:StudyUnit/pi:PhysicalInstance/@id
21 variable Study Variable /ddi:DDIInstance/s:StudyUnit//l:Variable/@id
22 product Study LogicalDataSet //s:StudyUnit/l:LogicalProduct/@id
23 owl:versionInfo Study
24 skos:definition Universe rdf:langString c:Universe/c:HumanReadable

notes

  • (2): inf code list is defined use it as the identifier
  • (9): the date the study is available to the public
  • (13): the URI to the DDI file(s) defined via param to the xslt
  • (21): suggested for identification

11.7.2 General Metadata

#
property
domain class
range class
mapping
1 adms:identifier disco:Study adms:Identifier /ddi:DDIInstance/s:StudyUnit/@id
2 adms:identifier disco:StudyGroup adms:Identifier
3 adms:identifier disco:AnalysisUnit adms:Identifier
4 adms:identifier disco:Universe adms:Identifier
5 adms:identifier disco:LogicalDataSet adms:Identifier
6 adms:identifier disco:DataFile adms:Identifier //pi:PhysicalInstance/pi:DataFileIdentification
7 adms:identifier disco:DescriptiveStatistics adms:Identifier
8 adms:identifier disco:SummaryStatistics adms:Identifier
9 adms:identifier disco:CategoryStatistics adms:Identifier
10 adms:identifier disco:Variable adms:Identifier //l:Variable/l:VariableName
11 adms:identifier disco:RepresentedVariable adms:Identifier
12 adms:identifier disco:Question adms:Identifier
13 adms:identifier disco:Instrument adms:Identifier
14 adms:identifier disco:Questionnaire adms:Identifier
15 skos_prefLabel rdfs:Resource rdf:langString
16 dcterms:relation rdfs:Resource foaf:Document
17 dcterms:description dcterms:RightsStatement rdf:langString
18 skos:prefLabel dcterms:RightsStatement rdf:langString
19 rdfs:seeAlso dcterms:RightsStatement foaf:Document
20 skos:prefLabel dcterms:PeriodOfTime rdf:langString
21 startDate dcterms:PeriodOfTime xsd:date
22 endDate dcterms:PeriodOfTime xsd:Date
23 skos:prefLabel dcterms:MediaTypeOrExtent rdf:langString
24 org:memberOf foaf:Person org:Organization

notes

  • (1): s:StudyUnit/r:Archive/a:ArchiveSpecific/a:Collection/a:CallNumber is also a candidate for identification

11.7.3 Data Sets, Data Files, and Descriptive Statistics

#
property
domain class
range class
mapping
1 instrument LogicalDataSet Instrument
2 dataFile LogicalDataSet DataFile
3 aggregation LogicalDataSet qb:DataSet
4 variable LogicalDataSet Variable
5 universe LogicalDataSet Universe
6 dcterms:title LogicalDataSet rdf:langString //l:LogicalProduct/r:Label
7 isPublic LogicalDataSet xsd:boolean
8 dcterms:accessRights LogicalDataSet dcterms:RightsStatement ancestor::s:StudyUnit/a:Archive/a:DefaultAccess/a:AccessConditions
9 dcterms:license LogicalDataSet dcterms:LicenseDocument
10 inputVariable qb:DataSet Variable
11 caseQuantity DataFile xsd:nonNegativeInteger //pi:PhysicalInstance/pi:GrossFileStructure/pi:CaseQuantity
12 dcterms:description DataFile rdf:langstring
13 owl:versioninfo DataFile string //pi:PhysicalInstance/@version
14 dcterms:temporal DataFile dcterms:PeriodOfTime
15 dcterms:spatial DataFile dcterms:Location pi:PhysicalInstance/r:Coverage/r:SpatialCoverage/@id | pi:PhysicalInstance/r:Coverage/r:SpatialCoverageReference/r:ID
16 dcterms:provenance DataFile dcterms:ProvenanceStatement
17 dcterms:subject DataFile skos:Concept
18 dcterms:format DataFile dcterms:MediaTypeOrExtend
19 statisticsDataFile DescriptiveStatistics DataFile
20 statisticsVariable SummaryStatistics Variable
21 invalidcases SummaryStatistics xsd:nonNegativeInteger
22 maximum SummaryStatistics xsd:decimal
23 mean SummaryStatistics xsd:decimal
24 median SummaryStatistics xsd:decimal
25 minimum SummaryStatistics xsd:decimal
26 mode SummaryStatistics xsd:decimal
27 standardDeviation SummaryStatistics xsd:decimal
28 validCases SummaryStatistics xsd:nonNegativeInteger
29 weightedInvalidCases SummaryStatistics xsd:nonNegativeInteger
30 weightedMean SummaryStatistics xsd:decimal
31 weightedMedian SummaryStatistics xsd:decimal
32 weightedMode SummaryStatistics xsd:decimal
33 weightedValidCases SummaryStatistics xsd:nonNegativeInteger
34 statisticsCategory CategoryStatistics skos:Concept
35 cumulativePercentage CategoryStatistics xsd:decimal
36 frequency CategoryStatistics xsd:nonNegativeInteger
37 percentage CategoryStatistics xsd:decimal
38 weightedCumulativePercentage CategoryStatistics xsd:decimal
39 weightedFrequency CategoryStatistics xsd:nonNegativeInteger
40 weightedPercentage CategoryStatistics xsd:decimal

notes

  • (7): not populated from DDI (could be set as an param to the xslt)
  • (17): located in pi:PhysicalInstance/r:Coverage/r:TopicalCoverage (both subject and keyword)

11.7.4 Variables, Variable Definitions, Representations, and Concepts

#
property
domain class
range class
mapping
1 skos:inScheme skos:Concept skos:ConceptScheme
2 skos:hasTopConcept skos:ConceptScheme skos:Concept
3 skos:broader skos:Concept skos:Concept c:Universe/c:SubUniverse/@id
4 skos:narrower skos:Concept skos:Concept
5 skos:definition skos:Concept rdf:langString c:Universe/c:UniverseName
6 skos:notation skos:Concept rdfs:Literal c:Universe/c:MachineReadable [skos:notation is only used to represent codes]
7 skos:prefLabel skos:Concept rdf:LangString c:Universe/r:Label [skos:notation is only used to represent categories]
8 question Variable Question //l:Variable/r:QuestionReference/r:ID
9 universe Variable Universe //l:Variable/r:UniverseReference/r:ID
10 analysisUnit Variable AnalysisUnit
11 concept Variable skos:Concept //l:Variable/r:ConceptReference/r:ID
12 representation Variable Representation
13 basedOn Variable RepresentedVariable
14 dcterms:description Variable rdf:langString //l:Variable/r:Description
15 skos:notation Variable rdfs:Literal //l:Variable/l:VariableName
16 skos:prefLabel Variable rdf:langString //l:Variable/r:Label
17 concept RepresentedVariable skos:Concept
18 universe RepresentedVariable Universe
19 representation RepresentedVariable Representation
20 dcterms:description RepresentedVariable rdf:langString
21 skos:prefLabel RepresentedVariable rdf:langString

notes

  • (12): not sure where to map to in DDI 3.1
  • (13): coming in DDI 3.2

11.7.5 Data Collection

#
property
domain class
range class
mapping
1 universe Question Universe //l:Variable/r:UniverseReference/r:ID
2 concept Question skos:Concept //l:Variable/r:ConceptReference/r:ID
3 responseDomain Question Representation
4 questionText Question rdf:langString //d:QuestionItem | d:MultipleQuestionItem/d:QuestionText/d:LiteralText/d:Text
5 skos:prefLabel Question rdf:langString //d:QuestionItem/d:QuestionItemName | d:MultipleQuestionItem/d:MultipleQuestionItemName
6 question Questionnaire Question
7 collectionMode Questionnaire skos:Concept
8 externalDocumentation Instrument foaf:Document
9 dcterms:description Instrument rdf:langString d:Intrument/r:Description
10 skos:prefLabel Instrument rdf:langString d:Instrument/r:Label

notes

  • (4): question-text exists for multiple elements
  • (5): the question name as label

12. Reference Implementations

12.1 Microdata Information System (MISSY)

The Microdata Information System (MISSY) is an online service platform that provides systematically structured metadata for official statistics. This includes data documentation at the study and variable level (6 series, 73 studies, 121 data sets, 22,719 variables, and 6,481 questions) as well as documentation materials, tools, and further information. We developed

  1. an editor in compliance with DDI-Codebook, DDI-Lifecycle, and Disco to improve and simplify the process of documentation and
  2. a web information system to provide the end user with various views on the metadata.

12.1.1 Data Models

We use Disco as core data model and extend it with a project-specific data model as Disco does not meet all of our project requirements. We provide open-source reference implementations of the Disco and the project-specific data model in Java, see Software Resources below. As instances of these data models may be physically stored in multiple formats such as DDI-XML, Disco, relational databases, and Java, we offer persistence implementations for each of these models according to their individual persistence APIs. Diverse export routines (e.g., Disco and DDI-Lifecycle) are available to enable the reuse of metadata in other systems.

12.1.2 Publications

12.1.3 Software Resources

13. Limitations

Disco provides a detailed structure to describe data for discovery purposes in the Semantic Web. This way, DDI XML repositories can be transformed to Disco and provided to the Linked Data Web. For other purposes like preservation, exchange, replication or metadata-driven approach, the DDI XML specifications Lifecycle and also Codebook provide a richer structure. Examples are advanced missing value description and the notion of a conceptual variable (derived from GSIM – Generic Statistical Information Model).

A. Vocabulary Reference

1. Studies and StudyGroups

Class: disco:Study
A Study represents the process by which a data set was generated or collected.
Object Property: disco:variable (Domain:disco:Study -> Range: disco:Variable )
Indicates the Variable of a Study.
Object Property: disco:inGroup (Domain:disco:Study -> Range: disco:StudyGroup )
points from a Study to the StudyGroup which contains the Study.
Object Property: disco:product (Domain:disco:Study -> Range: http://purl.org/linked-data/cube#LogicalDataSet )
Indicates the LogicalDataSets of a Studies.
Class: disco:StudyGroup
In some cases, where data collection is cyclic or on-going, data sets may be released as a StudyGroup, where each cycle or wave of the data collection activity produces one or more data sets. This is typical for longitudinal studies, panel studies, and other types of series (to use the DDI term). In this case, a number of Study objects would be collected into a single StudyGroup.
Class: disco:AnalysisUnit Sub Class of: skos:Concept
The process collecting data is focusing on the analysis of a particular type of subject. If, for example, the adult population of Finland is being studied, the AnalysisUnit would be individuals or persons.
Class: disco:Universe Sub Class of: skos:Concept
A Universe is the total membership or population of a defined class of people, objects or events.

2. Data Sets, Data Files, and Descriptive Statistics

Class: disco:LogicalDataSet Sub Class of: http://www.w3.org/ns/dcat#Dataset
Each study has a set of logical metadata associated with the processing of data, at the time of collection or later during cleaning, and re-coding. LogicalDataSet represents the microdata dataset.
Object Property: disco:variable (Domain:disco:LogicalDataSet -> Range: disco:Variable )
points to Variable contained in the LogicalDataSet
Object Property: disco:aggregation (Domain:disco:LogicalDataSet -> Range: http://purl.org/linked-data/cube#DataSet )
points to the aggregated data set of a microdata data set.
Datatype Property: disco:isPublic (Domain:disco:LogicalDataSet -> Range: xsd:boolean )
The value true indicates that the dataset can be accessed (usually downloaded) by anyone.
Datatype Property: disco:variableQuantity (Domain:disco:LogicalDataSet -> Range: xsd:nonNegativeInteger )
This property can be used when (1) no variable level information is available and when (2) only a stub of the RDF is requested e.g when returning basic information on a study of file, no information on potentially hundreds or thousands of variables references or metadata has to be returned.
Class: disco:DataFile Sub Class of: http://www.w3.org/ns/dcat#Distribution
The class DataFile, which is also a dcmitype:Dataset, represents all the data files containing the microdata datasets.
Datatype Property: disco:caseQuantity (Domain:disco:DataFile -> Range: xsd:nonNegativeInteger )
case quantity of a DataFile.
Datatype Property: disco:variableQuantity (Domain:disco:DataFile -> Range: xsd:nonNegativeInteger )
This property can be used when (1) no variable level information is available and when (2) only a stub of the RDF is requested e.g when returning basic information on a study of file, no information on potentially hundreds or thousands of variables references or metadata has to be returned.
Class: disco:DescriptiveStatistics
SummaryStatistics pointing to variables and CategoryStatistics pointing to categories and codes are both DescriptiveStatistics. Summary statistics are used to summarize a set of observations, in order to communicate the largest amount of information as simply as possible. A category statistic or frequency is the value of a statistic associated with a category value (even it can be applied to numeric values of metric variables). A frequency is the number of times a data value occurs. There are frequency counts (absolute) and percentages (relative) of the values of individual variables. See also the Wikipedia entry on frequency in statistics.
Object Property: disco:statisticsDataFile (Domain:disco:DescriptiveStatistics -> Range: disco:DataFile )
Indicates the DataFile of a specific DesciptiveStatistics individual.
Class: disco:SummaryStatistics Sub Class of: disco:DescriptiveStatistics
For SummaryStatistics, maximum values, minimum values, and standard deviations can be defined.
Object Property: disco:statisticsVariable (Domain:disco:SummaryStatistics -> Range: disco:Variable )
Indicates the Variable of a specific SummaryStatistics individual.
Object Property: disco:summaryStatisticsType (Domain:disco:SummaryStatistics -> Range: skos:Concept )
summary statistics type
Object Property: disco:weightedBy (Domain:disco:SummaryStatistics -> Range: disco:Variable )
Defines the weight variable of a category or summary statistic computation respectively value. It can also be used to indicate if a weight variable is used but the related variable is not known. weightedBy may be assigned to a category statistic value or to a summary statistic value.
Class: disco:CategoryStatistics Sub Class of: disco:DescriptiveStatistics
For CategoryStatistics, frequencies, percentages, and weighted percentages can be defined.
Object Property: disco:statisticsCategory (Domain:disco:CategoryStatistics -> Range: skos:Concept )
Indicates the skos:Concept (representing codes and categories) of a specific CategoryStatistics individual.
Object Property: disco:weightedBy (Domain:disco:CategoryStatistics -> Range: disco:Variable )
Defines the weight variable of a category or summary statistic computation respectively value. It can also be used to indicate if a weight variable is used but the related variable is not known. weightedBy may be assigned to a category statistic value or to a summary statistic value.
Datatype Property: disco:frequency (Domain:disco:CategoryStatistics -> Range: xsd:nonNegativeInteger )
frequency
Datatype Property: disco:percentage (Domain:disco:CategoryStatistics -> Range: xsd:decimal )
percentage
Datatype Property: disco:computationBase (Domain:disco:CategoryStatistics -> Range: rdf:langString )
computation base
Datatype Property: disco:cumulativePercentage (Domain:disco:CategoryStatistics -> Range: xsd:decimal )
cumulative percentage

3. Variables, Variable Definitions, Representations, and Concepts

Class: disco:Representation
The Representation of a variable is the combination of a value domain, datatype, and, if necessary, a unit of measure or a character set. Representation is one of a set of values to which a numerical measure or a category from a classification can be assigned (e.g. income, age, and sex: male coded as 1).
Class: disco:RepresentedVariable
RepresentedVariables encompasse study-independent, re-usable parts of variables like occupation classification. The Representation of a variable is the combination of a value domain, datatype, and, if necessary, a unit of measure or a character set. Representation is one of a set of values to which a numerical measure or a category from a classification can be assigned (e.g. income, age, and sex: male coded as 1).
Class: disco:Variable
Variables provide a definition of the column in a rectangular data file. Variable is a characteristic of a unit being observed. A variable might be the answer of a question, have an administrative source, or be derived from other variables.
Object Property: disco:basedOn (Domain:disco:Variable -> Range: disco:RepresentedVariable )
points to the RepresentedVariable the Variable is based on.

4. Data Collection

Class: disco:Question
A Question is designed to get information upon a subject, or sequence of subjects, from a respondent.
Object Property: disco:responseDomain (Domain:disco:Question -> Range: disco:Representation )
The response domain of questions.
Datatype Property: disco:questionText (Domain:disco:Question -> Range: rdf:langString )
question text
Class: disco:Instrument
The data for the study are collected by an Instrument. The purpose of an Instrument, i.e. an interview, a questionnaire or another entity used as a means of data collection, is in the case of a survey to record the flow of a questionnaire, its use of questions, and additional component parts. A questionnaire contains a flow of questions.
Object Property: disco:externalDocumentation (Domain:disco:Instrument -> Range: foaf:Document )
points from an Instrument to a foaf:Document which is the external documentation of the Instrument.
Class: disco:Questionnaire Sub Class of: disco:Instrument
A questionnaire contains a flow of questions.
Object Property: disco:collectionMode (Domain:disco:Questionnaire -> Range: skos:Concept )
mode of collection of a Questionnaire

5. Other properties

Class: disco:Question
A Question is designed to get information upon a subject, or sequence of subjects, from a respondent.
Object Property: disco:universe (Domain:disco:Study, disco:StudyGroup, disco:RepresentedVariable, disco:Variable, disco:Question, disco:LogicalDataSet -> Range: disco:Universe )
Indicates the Universe(s) of Studies, StudyGrous, RepresentedVariables, Variables, Questions, and LogicalDataSets.
Object Property: disco:concept (Domain:disco:RepresentedVariable, disco:Question, disco:Variable -> Range: skos:Concept )
points to the DDI concept of a RepresentedVariable, a Variable, or a Question
Datatype Property: disco:questionText (Domain:disco:Question -> Range: rdf:langString )
question text
Class: disco:Instrument
The data for the study are collected by an Instrument. The purpose of an Instrument, i.e. an interview, a questionnaire or another entity used as a means of data collection, is in the case of a survey to record the flow of a questionnaire, its use of questions, and additional component parts. A questionnaire contains a flow of questions.
Object Property: disco:externalDocumentation (Domain:disco:Instrument -> Range: foaf:Document )
points from an Instrument to a foaf:Document which is the external documentation of the Instrument.
Class: disco:Questionnaire Sub Class of: disco:Instrument
A questionnaire contains a flow of questions.
Object Property: disco:collectionMode (Domain:disco:Questionnaire -> Range: skos:Concept )
mode of collection of a Questionnaire
Object Property: disco:question (Domain:disco:Variable, disco:Questionnaire -> Range: disco:Question )
Indicates the Questions associated to Variables or contained in Questionnaires.

B. Combined UML Diagram

The following figure shows the object properties between the most important classes of the DDI-RDF Discovery Vocabulary. Additionally, the cardinalities of these object properties and class hierarchies are visualized.

Combined UML Diagram (object properties only)
Figure 37 Combined UML Diagram (object properties only)

A scalable version of this diagram can be found here.

C. Use Cases and Example Queries

Vompras, Gregory, Bosch, Capadisli, and Wackerow [Scenarios] have written a paper describing typical use cases associated with the DDI-RDF Discovery Vocabulary. The specification the DDI-RDF Discovery Vocabulary does not contain the full list of all the possible use cases. The complete list can be found in the mentioned paper. We now show a couple of representative use cases associated with the DDI-RDF Discovery Vocabulary.

Searching for subjects and temporal coverage

Find studies from years 2000 and after about climate change.

SELECT ?studyTitle ?studyAbstract ?logicalDataSetTitle
WHERE {
  ?study a disco:Study ;
    dcterms:title ?studyTitle ;
    dcterms:abstract ?studyAbstract ;
    dcterms:subject [ skos:prefLabel “Climate Change” ] ;
    dcterms:temporal [ disco:startDate ?date ] ;
    disco:product ?logicalDataSet .

  ?logicalDataSet a disco:LogicalDataSet ;
    dcterms:title ?logicalDataSetTitle .

  FILTER (?date >= 2000)
}

Searching for particular access conditions and rights

Find titles of data sets which are publicly available under the Canadian Data Liberation Initiative Community policy. Optionally give links to the rights statement and the license.

SELECT ?logicalDataSetTitle
WHERE {
  ?logicalDataSet a disco:LogicalDataSet ;
    dcterms:title ?logicalDataSetTitle ;
    disco:isPublic ?isPublic ;
    dcterms:accessRights ?rightsStatement .

  ?rightsStatement skos:prefLabel ?rightsStatementLabel .

  FILTER (
    ?isPublic = "true" &&
    ?rightsStatementLabel = "Data Liberation Initiative Community"
  )
  
  OPTIONAL {
    ?rightsStatement rdfs:seeAlso ?rightsStatementURL .
  }
  OPTIONAL {
    ?logicalDataSet dcterms:license ?licenseDocument .
  }
}

Searching for particular questions

Find all studies with questions about commuting to work.

SELECT ?studyTitle ?studyAbstract
WHERE {
  ?study a disco:Study ;
    disco:instrument ?questionnaire ;
    dcterms:title ?studyTitle ;
    dcterms:abstract ?studyAbstract .
  ?questionnaire disco:question ?question .
  ?question disco:questionText ?questionText .

  FILTER (regex(?questionText, "commut.*work"))
}

Searching for particular variables

Find study groups where the study uses the species variable and has a variable defined as Bufo alvarius

SELECT ?studyGroupTitle ?studyGroupAbstract
WHERE {
  ?study a disco:Study ;
    disco:inGroup ?studyGroup ;
    disco:variable ?variable .

  ?studyGroup dcterms:title ?studyGroupTitle .
  ?studyGroup dcterms:abstract ?studyGroupAbstract .

  ?variable disco:concept ?variableConcept .
  FILTER (regex(?variableConcept, "species", "i"))

  ?variable disco:basedOn ?representedVariable .
  ?representedVariable disco:concept ?representedVariableConcept .
  FILTER (regex(?representedVariableConcept, "Bufo alvarius", "i"))
}

Representing relationships between persons, organizations and datasets

Within the context of Disco, we reuse other well elaborated and accepted vocabularies as often as possible and reasonable. DCMI, FOAF, ORG, ADMS, and PROV-O build one block of complementary vocabularies. Their use is shown in one combined use case. DCMI is used in order to describe general metadata, FOAF and ORG are used to describe persons and organizations, we use ADMS for the persistent identification of objects like persons and organizations, and PROV-O is used to provide provenance information. A typical scenario within the social sciences community could be the following one:

ddi:EuropeanStudy
  a disco:Study;
  disco:product ddi:EuropeanDataSet;
  disco:fundedBy ddi:GESIS;

ddi:John 
  a foaf:Person;
  a prov:Agent;
  adms:identifier [ a adms:Identifier ];
  prov:wasAssociatedWith ddi:AggregationActivity;
  prov:actedOnBehalfOf ddi:DERI;
  org:memberOf ddi:GESIS.
  
ddi:EuropeanDataSet
  a disco:LogicalDataSet;
  a prov:Entity;
  disco:aggregation ddi:AggregatedEuropeanDataSet.
  
ddi:AggregatedEuropeanDataSet
  a qb:DataSet;
  a prov:Entity.
  
ddi:AggregationActivity
  a prov:Activity;
  prov:used ddi:EuropeanDataSet;
  prov:wasGeneratedBy ddi:AggregatedEuropeanDataSet;
  
ddi:DERI
  a prov:Agent;
  a org:Organization;
  adms:identifier [ a adms:Identifier ].
  
ddi:GESIS
  a org:Organization;
  adms:identifier [ a adms:Identifier ].
  
-----

SELECT 
  ?person
WHERE
{
  ?person rdf:type foaf:Person.
  ?person org:memberOf ?gesis.
  ?gesis a org:Organization.
  ?allbus a disco:StudyGroup.
  ?allbus dcterms:creator ?person.
}

-----

SELECT
  ?organization ?person
WHERE
{
  ?organization rdf:type org:Organization.
  ?person rdf:type foaf:Person.
  ?euSILC rdf:type disco:Study.
  {?euSILC dcterms:contributor ?person}
  UNION
  {?euSILC dcterms:contributor ?organization}
}

-----

SELECT
  ?identifierOrganization ?identifierPerson
WHERE
{
  ?organization rdf:type org:Organization.
  ?orgnization rdf:type foaf:Agent.
  ?organization adms:identifier ?identifierOrganization.
  ?person rdf:type foaf:Person.
  ?person rdf:type foaf:Agent.
  ?person adms:identifier ?identifierPerson.
  ?euLFS rdf:type disco:Study.
  {?euLFS dcterms:publisher ?person}
  UNION
  {?euLFS dcterms:publisher ?organization}
}

Representing datasets using specific statistical classifications

XKOS extends SKOS with two main objectives: the first one is to allow the description of statistical classifications, the second one is to introduce refinements of the semantic properties defined in SKOS. The semantic properties extend the possible relations that can be applied between pairs of skos:Concepts. SKOS allows the following relations: skos:broader than, skos:narrower than, and skos:related to. The first two are hierarchical relations, one in each direction. In Disco, these SKOS properties may be substituted by additional XKOS properties like xkos:generalizes, xkos:hasPart, xkos:caused, xkos:previous, and xkos:next.

One question, typically asked by social science researchers, could be to query all the datasets (disco:LogicalDataSet) which have a specific statistical classification (skos:ConceptScheme) like ISCO (International Standard Classification of Occupations) or ANZSIC (Australian and New Zealand Industry Classification). It is also possible to query on the semantic relationships which are defined for statistical classifications using XKOS properties. By means of these properties not only hierarchical relations can be queried but also for example part of relationships (xkos:hasPart), more general (xkos:generalizes) and more specific (xkos:specializes) concepts, and positions of concepts in lists (xkos:previous, xkos:next).

The following figure gives an example inspired by the ANZSIC (Australian and New Zealand Industry Classification), which is a classification covering the field of economic activity. A small excerpt is shown here, limited to the classification object itself and its levels, as well as one item of the most detailed level (Class 6720 – Real Estate Services) and its parent items. Note that the URI employed in this example are entirely fictitious, since the ANZSIC has not yet been published as RDF.

For clarity, the properties of the classification items (code, labels, notes) have not been included in the figure.

Figure 38 Statistical classification – ANZSIC

On the left of the figure is the skos:ConceptScheme instance that corresponds to the ANZIC 2006 classification scheme, with its various SKOS and Dublin Core properties. Additionnal XKOS properties indicate that the classification has four levels and covers the field of economic activity, represented here as a concept from the EuroVoc thesaurus. In this case, the coverage is intended to be exhaustive and without overlap, so xkos:coversExhaustively and xkos:coversMutuallyExclusively could have been used together instead of xkos:covers.

The four levels are instances of xkos:ClassificationLevel; they are organized as a rdf:List which is attached to the classification by the xkos:levels property. Some level information has been represented on the top level, for example its depth in the classification (xkos:depth) and the concept that characterizes the items it is composed of (xkos:organizedBy). In the same fashion, concepts of subdivision, group and class could be created to describe the items of the lower levels.

The usual SKOS properties are used to connect the classification items to their respective level (skos:member) and to the classification (skos:inScheme or its specialization skos:topConceptOf) for the items of the first level). Similarly, skos:narrower is used to express the hierarchical relations between the items, but the subproperties defined in this specification could also be used. For example, xkos:hasPart could express the partitive relation between subdivision 67 ("Property Operators and Real Estate Services") and group 672 ("Real Estate Services").

Representing relationships between datasets, collections and data catalogs

While Disco and Data Cube provide terms for the description of datasets, both on a different level of aggregation, DCAT enables the representation of these datasets inside of data collections like repositories, catalogs or archives. The relationship between data collections and their contained datasets is useful, since such collections are a typical entry point when searching for data.

A search for data may consist of two phases. In a first phase, the user searches for different records described by dcat:CatalogRecord inside a data catalog. This search can differ according to the users’ information need. While it is possible to search for metadata provided inside such a record like dcterms:title, dcterms:description, etc., the user can also formulate a query to search for more detailed information about the dataset (represented as dcat:Dataset) or its distribution (dcat:Distribution), which are part of the record. For example, a user may want to search for datasets covering a particular topic (dcat:keyword), particular temporal and spatial coverages (dcterms:temporal and dcterms:spatial), or particular formats in which a distribution of the data is available (dcterms:format). Instances of dcat:DataSet are also described by specific themes they cover (dcat:theme). Since these themes are organized in a theme taxonomy (implemented by a skos:ConceptScheme and classes of skos:Concept), these themes can also be used for an overall search in all datasets of the data catalog.

Nevertheless, the search of the first phase will result in one or presumably multiple hits of datasets. Hence, another search has to be executed in a second phase in order to find out which datasets are relevant for the user, e.g. particular universes or samples. The search regarding particular criteria in multiple Disco datasets materializes as those described in the previous two use case sections and those presented in [9]. However, the user may find data sets which are published in Data Cube. In order to discover the original microdata source of a qb:DataSet, the property prov:wasDerivedFrom can hold the link the particular DDI data set disco:Study.

A user searching for data regarding dissatisfaction with politics in Europe may find the records :EuropeanStudy and :AggregatedEuropeanData in a :DataCatalog. By analyzing the information given in the themes and keywords of the associated data sets, the user can decide which data set is best suitable for his information need. He notices also that :AggregatedEuropeanDataset has been derived from :EuropeanDataset and seems to cover only a subset of the microdata set. If he is interested in the microdata instead of aggregated data, he is thus able to find the underlying microdata set.

ddi:DataCatalog_1
a dcat:Catalog;
dcat:record ddi:EuropeanStudy;
dcat:record ddi:AggregatedEuropeanData;
dcat:dataset ddi:EuropeanDataset;
dcat:dataset ddi:AggregatedEuropeanDataset.
          
ddi:EuropeanStudy
a dcat:CatalogRecord;
a disco:Study;
foaf:primaryTopic ddi:EuropeanDataset;
disco:product ddi:EuropeanDataset.
          
ddi:AggregatedEuropeanData;
a dcat:CatalogRecord;
foaf:primaryTopic ddi:AggregatedEuropeanDataset.
          
ddi:EuropeanDataset
a dcat:Dataset;
a disco:LogicalDataSet;
dcat:theme ddi:topics/WellBeing;
dcat:theme ddi:topics/PoliticalAttitudes;
dcat:keyword "Europe"@en;
dcat:keyword "Politics"@en.
          
ddi:AggregatedEuropeanDataset
a dcat:Dataset;
a qb:DataSet;
dcat:theme ddi:topics/PoliticalDissatisfaction;
dcat:keyword "Europe"@en;
dcat:keyword "Politics"@en;
prov:wasDerivedFrom ddi:EuropeanStudy.

D. Acknowledgements

This work has been started at the first workshop on “Semantic Statistics for Social, Behavioural, and Economic Sciences: Leveraging the DDI Model for the Linked Data Web” at Schloss Dagstuhl - Leibniz Center for Informatics, Germany in September 2011 organized by Richard Cyganiak, Arofan Gregory, Wendy Thomas, and Joachim Wackerow. This work has been continued at these three meetings:

This work has been supported by contributions of the participants of the events mentioned above:

We would like to thank the following organizations which have supported this work:

E. References

E.1 Normative references

[DCAT]
Data Catalog Vocabulary (DCAT), http://www.w3.org/TR/vocab-dcat/
[DCMI]
DCMI Metadata Terms (DCMI), http://dublincore.org/documents/dcmi-terms/
[FOAF]
Friend of a Friend (FOAF), http://www.foaf-project.org/
[FundRef]
FundRef, http://www.crossref.org/fundref/
[ORCID]
ORCID, http://orcid.org/
[ORG]
Organization Ontology (ORG), http://www.w3.org/TR/vocab-org/
[PROV-O]
PROV Ontology (PROV-O), http://www.w3.org/TR/prov-o/
[RDF Data Cube Vocabulary]
RDF Data Cube Vocabulary, http://www.w3.org/TR/vocab-data-cube/
[SKOS]
Simple Knowledge Organization System (SKOS), http://www.w3.org/2004/02/skos/
[SIO]
Semanticscience Integrated Ontology, https://code.google.com/p/semanticscience/wiki/SIO

E.2 Informative references

[ADMS]
Asset Description Metadata Schema (ADMS), http://www.w3.org/TR/vocab-adms/
[CoRR-2014]
Bosch, T., Wira-Alam, A., & Mathiak, B. Designing an Ontology for the Data Documentation Initiative. 2014. Computing Research Repository (CoRR). URL: http://arxiv.org/abs/1402.3470
[DDI-RDF-Discovery-Vocabulary]
Bosch, T., Cyganiak, R., Gregory, A., Wackerow, J. DDI-RDF Discovery Vocabulary: A Metadata Vocabulary for Documenting Research and Survey Data. 2013. Proceedings of the WWW2013 Workshop on Linked Data on the Web. URL: http://ceur-ws.org/Vol-996/papers/ldow2013-paper-12.pdf
[DDI-Working-Paper-Series-2012]
Block, W., Bosch, T., Fitzpatrick, B., Gillman, D., Greenfield, J., Gregory, A., Hebing, M., Hoyle, L., Humphrey, C., Johnson, J., Linnerud, J., Mathiak, B., McEachern, S., Radler, B., Risnes, Ø., Smith, D., Thomas, W., Wackerow, J., Wegener, D., Zenk-Möltgen, W. Developing a Model-Driven DDI Specification. 2012. DDI Working Paper Series. URL: http://www.ddialliance.org/resources/publications/working-papers
[ESWC-2011]
Bosch, T., Wira-Alam, A., Mathiak, B. Designing an Ontology for the Data Documentation Initiative. 2011. Proceedings of the 8th Extended Semantic Web Conference (ESWC), Poster-Session. URL: http://www.eswc2011.org/content/accepted-posters.html
[eXtended Knowledge Organization System]
Dan Gillman, Franck Cotton, and Yves Jaques eXtended Knowledge Organization System (XKOS). 2013. METIS, Work Session on Statistical Metadata. URL: http://www.unece.org/stats/documents/2013.05.metis.html
[IASSIST-Quarterly-38(4)-39(1)_1]
Bosch, T., Mathiak, B. Use Cases Related to an Ontology of the Data Documentation Initiative. 2015. IASSIST Quarterly, 38(4) & 39(1). URL: http://iassistdata.org/iq/issue/38/4
[IASSIST-Quarterly-38(4)-39(1)_2]
Bosch, T., Olsson, O., Gregory, A., & Wackerow, J. DDI-RDF Discovery - A Discovery Model for Microdata. 2015. IASSIST Quarterly, 38(4) & 39(1). URL: http://iassistdata.org/iq/issue/38/4
[IASSIST-Quarterly-38(4)-39(1)_3]
Bosch, T. & Zapilko, B. Semantic Web Applications for the Social Sciences. 2015. IASSIST Quarterly, 38(4) & 39(1). URL: http://iassistdata.org/iq/issue/38/4
[IASSIST-Quarterly-38(4)-39(1)_4]
Schaible, J., Zapilko, B., Bosch, T., & Zenk-Möltgen, W. Linking Study Descriptions to the Linked Open Data Cloud. 2015. IASSIST Quarterly, 38(4) & 39(1). URL: http://iassistdata.org/iq/issue/38/4
[Linked-Statistical-Data]
Bosch, T., Cyganiak, R., Wackerow, J., and Zapilko, B. Leveraging the DDI Model for Linked Statistical Data in the Social, Behavioural, and Economic Sciences. 2012. International Conference on Dublin Core and Metadata Applications. URL: http://dcpapers.dublincore.org/pubs/article/view/3654
[Scenarios]
Vompras, J., Gregory, A., Bosch, T., Capadisli, and Wackerow, J.Scenarios for the DDI-RDF Discovery Vocabulary. May 2013. DDI Working Paper Series – Semantic Web 2. DOI: 10.3886/DDISemanticWeb02
[Semantic Statistics]
Cyganiak, R., Field, S., Gregory, A., Halb, W., Tennison, J. Semantic Statistics: Bringing Together SDMX and SCOVO. 2010. Proceedings of the WWW2010 Workshop on Linked Data on the Web. URL: http://ceur-ws.org/Vol-628/ldow2010_paper03.pdf
[SemStats-2013]
Bosch, T., Zapilko, B., Wackerow, J., & Gregory, A. Towards the Discovery of Person-Level Data - Reuse of Vocabularies and Related Use Cases. 2013. Proceedings of the 1st International Workshop on Semantic Statistics (SemStats 2013), 12th International Semantic Web Conference (ISWC 2013). URL: http://semstats.github.io/2013/proceedings
[XKOS]
SKOS Extension for Statistics (XKOS), http://htmlpreview.github.io/?https://github.com/linked-statistics/xkos/blob/master/xkos.html