[DDI-SRG] Referencing codes and dimensions

[Joachim Wackerow]

Actually my intention was to ask for similar identification 
constructions, the reference from category statistics has the target 
code in codescheme.

These identification constructions are independent from variable.

Achim

Wendy Thomas wrote:
> The other place this type of reference is used is in category 
> statistics. The thing is that since you are talking about a code scheme 
> value as it is used in a specific variable you would need (I think) to 
> have a new unique ID for each code value within the variable making for 
> a huge amount of additional coding (id for each code in the codeScheme 
> plus ID for each code in the Variable)....ugh
> 
> wendy
> 
> 
> On Thu, 17 Jan 2008, Joachim Wackerow wrote:
> 
>> Arofan,
>>
>> Looking at the inline ncube stuff and discussing these issues with Wendy
>> I realized the following:
>>
>> Codes in coding schemes and dimensions in NCubeLogicalProduct are
>> referenced not in the usual way. They have no identification but they
>> need to be referenced. The reference is indeed made by a combination of
>> the identification of an upper level container and a content value at
>> the specific element.
>>
>> I know these are not really new findings. Now I'm wondering if this
>> solution is really desirable, to have different methods of referencing.
>> At least there seems to be an inconsistency between code and dimension
>> (usage of element/attribute). Do we have similar references at other 
>> places?
>>
>> Code
>>
>> <l:CodeScheme>
>>   <l:Identifier>
>>     <r:ID>CodeS_2</r:ID>
>>   </l:Identifier>
>>   <l:CategorySchemeReference>
>>     <r:ID>CATS_1</r:ID>
>>   </l:CategorySchemeReference>
>>   <l:Code>
>>     <l:CategoryReference>
>>       <r:ID>C_1</r:ID>
>>     </l:CategoryReference>
>>     <l:Value>1</l:Value>
>>   </l:Code>
>>
>> The steps to find the code 1 would be:
>> 1. generate the URN for all Identifier (normalization)
>> 2. find CodeScheme with matching URN
>> 3. apply XPath expression l:Code[ l:Value = '1' ]
>>
>> Step 1 and 2 are the common steps to find an Identifier. Step 1 is done
>> probably only once for all Identifier, then an index should be provided
>> in the application.
>> Step 3 is an expensive test on content of an element. This should be
>> also done only once to provide an index.
>>
>>
>> Dimension
>>
>> <l:NCube dimensionCount="3" cellCount="40">
>>   <l:Identifier>
>>     <r:ID>NCube</r:ID>
>>   </l:Identifier>
>>   <!-- - - - - - -->
>>   <l:Dimension rank="1">
>>     <l:VariableReference>
>>       <r:ID>V1</r:ID>
>>     </l:VariableReference>
>>   </l:Dimension>
>>
>> The steps to find the dimension with rank 1 would be (required by ncube
>> modules):
>> 1. generate the URN for all Identifier (normalization)
>> 2. find NCube with matching URN
>> 3. apply XPath expression l:Dimension[ @rank = '1' ]
>>
>> This is similar as at code but the test is done on the value of an
>> attribute.
>>
>>
>> Would it be possible to construct an Identifier for Code and Dimension
>> where the ID part is the critical value (Value or @rank) and the rest of
>> the URN is inherited by the upper level container (CodeScheme or Ncube).
>> This way step 3 would be probably unnecessary and a more consistent way
>> of referencing over all would exist.
>>
>> Achim
>> _______________________________________________
>> DDI-SRG mailing list
>> DDI-SRG at icpsr.umich.edu
>> http://www.icpsr.umich.edu/mailman/listinfo/ddi-srg
>>
> 
> 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


-- 
GESIS - German Social Science Infrastructure Services
http://www.gesis.org/en/