[DDI-SRG] Mixing code and value ranges

[Wendy Thomas]

Pascal,

We have two options for providing a means of specifying labeled codes for 
missing values in numeric/text or any other non-codescheme representaiton.

1) allow multiple variabiable representations which opens the door to just 
about any combination for any reason

2) expand the definition information for missing value within 
representation to include the ability to reference a codescheme of missing 
values and/or declare the missing values specifically. Right now we can 
only do the second of these. If this is the approach taken then numeric 
ranges, text ranges, codeschemes etc used as representations should 
include only the valid non-missing responses. All missing values for each 
type should be declared separately within the representation under 
missingValues which should be expanded to allow ranges and reference to a 
codescheme of missing values. In this way the "valid" responses that can 
be validated in the data file is the set of non-missing respresentations 
PLUS the set of missing values. This clearly differentiates both types and 
does not double count (represent) them.

Wendy




On Fri, 11 Jul 2008, Pascal Heus wrote:

> Wendy:
> Thanks for the feedback. The problem is this doesn't work when you have a 
> numeric value that is a range (and not a coded list). The example of "Number 
> of cars you owned (99=don't know)" is not associate with a code/category. 
> 1,2,3,4,5,6,55 are all valid answers and I do not need to create a code list 
> (it is a numeric variable). Also, 99 is not particularly a missing value. I 
> could have other codes. What about "How much money did you earn last year (-1 
> or 9999999999 for do not want to disclose)? I know it's a bad way to code 
> this but you see this a lot in survey questions. I think we're being a bit 
> too restrictive here. We should be able to have both Code and Numeric/Text 
> representation. In this case, if it's not in the code list it is still. a 
> valid response.
> *P
>
> Wendy Thomas wrote:
>> Ok...right now you would list the valid range and list 99 in the attribute 
>> missingValue. We have on our list (from my memory..should check) a review 
>> of how we identify missing values in a consistant way for all 
>> representations. This should include the ability to reference a codescheme. 
>> By doing this we can differentiate between the valid set of non-missing 
>> data responses and missing data responses for each variable. It would also 
>> make it easier to reuse code schemes between questions and variables as the 
>> missing value codes could be declared in the variable representation rather 
>> than being affixed to each code scheme (render a new set of codes schemes 
>> that include the question codescheme plus the missing value codescheme.
>> 
>> Anyway...thats where we are. The reason for NOT making the representation 
>> repeatable is to prevent abuse. People want to put in both a codescheme 
>> representation AND a numeric representation. The comment was "how else will 
>> I know what the valid response numbers are?" the answer was if its not in 
>> the code scheme its not valid.
>> 
>> Ideally I think the representation description should list the valid data 
>> values and the valid missing values should be in missingData.
>> 
>> Wendy
>> 
>> 
>> On Thu, 10 Jul 2008, Pascal Heus wrote:
>> 
>>> Wendy and all:
>>> In DDI, how do we document a variable that mixes value ranges and
>>> code(s)? With ValueRepresentation no longer non-repeatable, this now
>>> seem tricky....
>>> For example, I can have a numeric variable like "Number of cars owned"
>>> with 99="don't know". I can I say this must be in the 0-99 range and
>>> link the value 99 to a code?
>>> I'm wondering if ValueRepresentation and CodeRepresentation should be
>>> separated or have the ability to have a CodeRepresentation and
>>> Date/Numeric/TextRepresentation at the same time.
>>> thanks
>>> *P
>>> _______________________________________________
>>> 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
>> 
>

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