[DDI-users] Group vs Universe

[Wendy Thomas]

Briefly, Group is used to put Study Units and SubGroups (groups of Study 
Units) into an organzed structure for ad-hoc comparison purposes or to 
represent the repetative nature of Study Unit such as in longitudinal or 
cross-sectional studies. This form of grouping-by-design supports 
inheritance. Inheritance is where say a survey is repeated yearly. The 
survey itself is inherited and the individual study units related study 
specific information such as dates, locations, etc.

Universe is a conceptual component of a study (or group of studies). The 
Study Unit references its universe which is housed in a Universe Scheme. A 
Universe Scheme can be specific to a study and the sub-universes reflected 
in the instrument flow (if then branches) or multiple sub-universes 
refelcted in how the study is administered (patients, hospitals, doctors).

I realize you're clear on the above but since this is a reply to the 
community I wanted to state this first.

Now to your 2 studies. You have a number of options depending on what 
activities you are trying to support.

1) If these were designed to be comparative, you can create a group, 
describe the Universe Scheme at the group level and reference from the 
study unit. The Universe Scheme itself provides the relationship between 
the Universes of both studies. For Example a top level Universe would be 
the more inclusive and the other a sub-universe. Now the universes you 
described have several hierachies in them (temporal, spatial, topical)

1970-2010, Families in the USA

Hierarchy 1: 1990-2010

Hierarchy 2: Families in USA with child/children diagnosed autistic within 
the spectrum

StudyUnit A would pount to the full Universe
StudyUnit B would point to the subuniverse in Hierarchy 1 AND the 
subuniverse in Hierachy 2

Note that if more than one universe is reference it is the OVERLAP of 
those two universes that is being defined not the combined area.

So with this information you can tell that both studies are within the 
same hierarchies but A has a broader coverage as StudyUnit B points to 
only subuniverses.

It is clear that StudyUnitA has data on families without autistic children 
but you would then need to check the variables and category schemes to see 
if you could identify those families within the data.

Your retrieval case (a) is absolutely clear in that the use of the same 
universe scheme clearly shows that the universe of  Study B is a subset of 
A. In case (b) the universe is a first step. Clearly STudy Unit B will not 
help you at all. Study Unit A will only help if it asked about the 
presence or absense of autistic children which can be used a base for 
filtering the cases. Alternatively if you can actually match cases between 
the two studies, for the years 1990-2010 you can match cases and use only 
those from Study A that do not have a match in study B.

If this was not a series and you are coming to it after the fact (say 
building a retrieval system) you can create a master Universe Scheme and 
then map the individual universe schemes to it. Your initial search would 
be on the master universe and then track back to all of the sources that 
reference that target. Again you will have the problem of identifying 
families without autistic children as it may be impossible to do case 
matching.

Your retrieval system would need to address how these two data sets were 
merged.

Wendy




On Thu, 27 Jan 2011, My Coyne wrote:

> Hi DDI Community,
>
> After reading and viewing the xsd for group and universe concepts, I find that I don't have a clear view of the these two
> concepts.  Would someone please  give me a better understand of both of these in comparison.
>
> 1.  I understand that group is used to group StudyUnits  and subgroups (assuming that subgroups are not study units)
>      -- Grouping factors (differentiators) can be: time, instrument, panel, geography, datasets, language
> 2.  I understand that Universe is an attribute of StudyUnits
>
> I would like to present a fictitious scenario in helping me solidifying the concepts.
>
> StudyUnit A
>      Universe A: all families of the USA
>     Timeframe:   1970 - 2010
>      In this study data collected are: marital status, education, incomes, geographical residence, religions, political affiliation for
> both parents and children
>
> StudyUnit B
>      Universe B: All families of the USA who have child/children diagnosed autistic within the spectrum
>      Timframe:   1990-2010
>      In this study similar data as in StudyUnit A are collected with additional information for autistic diagnosis and genotypes of
> the children.
>
> Using Grouping and/or Universe concepts, Could I then retrieve datasets that contains
> (a)  information that intersect of the two universes  (of studyA and StudyB)
> (b)  information about the families that do not have autistic children
> ?
>
> Thank you for your help.
>
>
>
>
> I would like to retrieve two datasets D1: with all children of the USA
>
> My Coyne
> mcoyne at boninc.com
> (cel): 301-399-6351
>
>
>
>
> _______________________________________________
> DDI-users mailing list
> DDI-users at icpsr.umich.edu
> http://www.icpsr.umich.edu/mailman/listinfo/ddi-users
>

Wendy L. Thomas                          Phone: +1 612.624.4389
Data Access Core Director		 Fax:   +1 612.626.8375
Minnesota Population Center              Email: wlt at pop.umn.edu
University of Minnesota
50 Willey Hall
225 19th Avenue South
Minneapolis, MN 55455