dct:hasPart and dct:isPartOf would be the appropiate properties.

Is this information readily available in existing metadata?

Can the owner of the issue provide an example for this information requirement? 

At Publications Office, there is as least two use cases:

- the WhoIsWho dataset of each institution could be declared as part of a global dataset which will represent the complete collection of the WhoIsWho at a given point of time

- the controlled vocabularies from MDR will be both distributed as individual datasets and included in a dataset package including all the vocabularies of a given release. It would be useful to related the individual datasets to the dataset describing the complete package. 

May I express my concern regarding this issue? Not that I disagree with the need of these properties but I would expect certain clarification to assure that is used undoubtedly. 

What exactly does it mean "a dataset which is made of several independent Datasets"?

if  <Ai> dct:isPartOf <B>

1. Would that mean that the entire dataset A is also included in B, in terms that any data found in A would also be in B?
Or would that mean that data in dataset A0, A1, etc together form B? In the later case, would it mean that B is an "abstract" dataset?

2. Is it expected that they have the same schema? Namely should they contain exactly the same information? What if A0 contains more information than A2 or even B?

3. And in respect to 2, should the relationship occur on Dataset level or on Distribution level (or both)?
 

This issue was also discussed during the 2nd GeoDCAT-AP call.

The background:

ISO 19115 (and INSPIRE) has the notion of "dataset series", which is generically defined as follows:

collection of datasets [...] sharing common characteristics.

Dataset series are commonly used in the geospatial domain, typically to model time series (but also other grouping criteria are used, as spatial coverage).

ISO 19115 models the relationship between dataset series and the resources in the collection by using constructs semantically analogous to dct:hasPart and dct:isPartOf. So, the mapping is pretty much straightforward. Notably, the resources in a dataset series can be datasets as well as other dataset series, thus enabling to organise the collection hierarchically.

Since DCAT /DCAT-AP do not support a specific class, the current GeoDCAT-AP draft specification proposes to model both dataset series as dcat:Dataset's, and, optionally, to add a dct:type statement, taking as object the relevant code value from the INSPIRE Registry:

http://inspire.ec.europa.eu/metadata-codelist/ResourceType/series/

I would also like to add that dataset series can group static as well as dynamic data. This especially when used for time series. In such a case, the collection can be increased other time, e.g., by adding data created / collected with the same criteria, but for new time periods.

Note that this does not result in a new version of the dataset / series. Actually, there may be an analogy with periodical publications. In other words, a dataset series can be seen as a journal, where articles (datasets) are published other time.

Librarians and the publishing industry have been addressing this issue for a long time. So, in case we want to model also this scenario, I wonder whether we can adopt / adapt their solutions. And probably Dublin Core already offers all what we need, considering that it was originally meant to describe bibliographic records. E.g., I guess that the dct:accrual* properties were specifically defined for periodicals.

Merged with https://joinup.ec.europa.eu/discussion/mo12-grouping-datasets

Merged with: https://joinup.ec.europa.eu/discussion/mo12-grouping-datasets