ADaM_Queries_and_Que..

advertisement
ADaM Queries and Questions that have arisen whilst building the Semantic RDF model.
1. The following list of groups of controlled terms is required in ADaM variables and yet not
present in the ADaM CT or SDTM CT:






Y,N,
Y,blank,
1,blank
BROAD/NARROW as it relates to SMQs
1,2 as it relates to SMQs
1,2,3 as it relates to AE severity
2. Most dates, times and datetimes are classified as “Numeric” since most people use SAS
practically these are set as SAS proprietary representations and so integers. The use of SAS
dates and times is not stated explicitly in the documentation ? Why is this ? and would there be
any other type of date/time representation ?
3. Can keys be legitimately expressed in the model ?
Primary keys are stated as “examples” in the documentation, should they not be more explicit
than that even as fixed keys at least for ADSL ? Would be better to have them defined for third
party tools/transformations ? However keys in practice are not that straightforward for example
4.2.1.6 in the IG also requires DTYPE to be part of the primary key ?
5. We discussed that flags should be called “Flags” or “Flagging Indicators” but not “Indicators” it
is felt “Indicator” does not match the most commonly used semantics in the document and real life.
6. Is “Analysis Descriptors “ an unnecessary semantic subheading of – DTYPE, Windowing TTE
and Labs ? if indeed it will be kept, does it add any meaning to the ADaM model and should it
therefore be expressed in RDF ?
7. Should the classifiers (roles, type and compliance) be part of the adam2.1 or adamig1.0 model ?
8. Taking a look at the TTE documentation and hence trying to model it, there are a problems
mainly because three variables are already defined in the BDS document. One problem is that
CNSR is Conditional in BDS and Required in TTE doc (in the note). Another problem is that the
BDS variables carried into the TTE guide of the same name can have different cdisc notes and
yet ADTTE is not an officially named dataset like ADAE is, it is a class of BDS. Is it not better just
to take TTE variables out of the BDS definition and make ADTTE another dataset type like ADAE ?
After all the basic primary key for TTE is different to BDS and it does say in the docs that it is
preferable not to mix TTE data with BDS data in the same BDS dataset. Anyway… the way TTE
is documented causes modeling problems. Do we make an ADTTEmodel/dataset showing what
is in the TTE guide ? or do we just add any extra TTE type variables to BDS ? if we add them to
BDS what do we do about compliance for CNSR and differing cdisc notes ?
Download