SDTM in RDF Questions

From PHUSE Wiki
Jump to: navigation, search

CDISC2RDF SDTM Ontology Issues/Questions

  • Explain CDISC2RDF modeling decisions. Why not use full ISO11179? How was the content chosen?

Content Issues/Questions

  • Representation of "add-on" permissible variables
  • Representing IG-specific domain assumptions (i.e. domain assumptions listed in the IG below each domain)
  • Representing general domain assumptions. Should some of these be represented with the model level?
  • Represent "instructional" text from the IG. For example, see the explanation of the different methods to relate PC and PP.
  • Should the model be a context for the IG?
  • Why aren’t the observation classes defined under mms:DatasetContext represented as classes instead of instances? Then the SDTM domains could be defined as instances of their particular observation class, as opposed to using mms:context. Given that the individual DatasetContexts represent different kinds of Datasets, it seems like it would make more sense to have the particular Datasets be subclasses of these DatasetContexts.

To me, there is a question of the semantics of mms:context. In this case, the context of a dataset refers to what kind of dataset it is. But when you look at mms:DataElements, their mms:context refers to the table they are associated with, which is a containment type of relationship, which is very different from an is_kind_of relationship.

  • The instances of mms:DataElement that are of the form similar to sdtm:DE.Event.—SCAT are essentially templates for variables (data elements). It does not make sense to me that these would be instances of the same class (mms:DataElement) as the variables themselves.
  • For the instances of mms:VariableGrouping, these are the categorization of variables based on the observation classes, which are instances of mms:DatasetContext. In most cases, but admittedly not in all cases, these are groupings of these template variables, which have the – at the beginning of their name. In this case, this is merely a grouping of variables with respect to variable categories, as opposed to a containment type relationship. Again, the question is what mms:context is supposed to represent. A grouping of some template variables seems like a different modeling situation than the containment of a set of variables in a dataset. Perhaps using subproperties of mms:context may make the abstractions clearer.