[DDI-SRG] [Fwd: Coded missing values]

[Joachim Wackerow]

Pascal and others,

It is possible to define a category without label, just stating the fact 
that the category is missing.

A missing value seems to be an attribute of a category not of a code. 
The term 'value' of missing value is here a bit misleading, it is not 
the 'Value' of 'Code'. A code is just a code. A category gives the code 
a meaning.

1.
A range of missing values seems to be convenient, but for my knowledge 
this kind of definition is only possible in SPSS. Nevertheless it is 
worth to think about.


Here are additional thoughts on missing values.

2.
I was wondering if it makes sense to define a missing value for a whole 
set of variables of for a whole study, which is common practice. I'm not 
sure if 'PhysicalDataProduct/MissingData' covers this sufficient.

3.
Common practice is also to define a blank as missing value for numeric 
variables. I.E. SPSS would assign a system missing value. Currently it 
is possible to define blank as a string code in Code/Value like:
'<Value> </Value>'. This seems to be error prone. A better approach 
would be to have an explicit definition for special codes like a blank. 
For example:
<Value blank="true"/>

4.
System missing values are not describable yet. But this seems to be 
important if a study is only available in SPSS or SAS. These values are 
represented as numeric values beyond the allowed real numbers internally 
in the statistical packages. Currently it would be only possible to 
recode the system missing value code to a valid numeric code described 
as missing category. This doesn't seem to be a good approach.

In SPSS one system missing value code is available, in SAS several are 
available. Again an explicit definition of this system missing value 
codes would be desirable. The name of the system missing value is 
important and in addition the related scheme like SAS. For example:
<Value missingValue="N" missingScheme="SAS"</>

I have the impression that we should discuss the subject "missing 
values" on a conference call (topic 1-4, perhaps other ideas). The 
existing possibility to define a missing value seems to be not sufficient.

Achim

Pascal Heus wrote:
> Further to the message below. in SPSS, I can also define a range of 
> coded missing values (like 10-20), is this something we'll support in 3.0?
> thanks
> *P
> 
> -------- Original Message --------
> Subject: 	Coded missing values
> Date: 	Wed, 29 Aug 2007 11:47:23 -0400
> From: 	Pascal Heus <pascal.heus at gmail.com>
> To: 	DDI Structural Reform Working Group. <ddi-srg at icpsr.umich.edu>
> 
> 
> 
> All:
> I have a couple of questions on coded missing values (like -1=missing, 
> not system missing) :
> - In DDI 2.0, missing values are flagged in the catgry element of the 
> variable. However, in SPSS, I can define missing value codes without 
> defining categories. In such case, is there a way to capture this in DDI 
> 2.0?
> - How do we capture missing value codes in DDI 3.0? According to Wendy's 
> mapping, it goes into the l:category element. As in the SPSS example 
> above, I could however define a discrete code value that is not 
> associated with a l:category but represent a missing value. Should I 
> file this issue in Mantis?
> thanks
> *P
> 
> -- 
> Imagination is more important than knowledge.
> - Albert Einstein
> www.quotator.net
> 
> 
> 
> -- 
> Do not dwell in the past, do not dream of the future, concentrate the mind on the present moment.
> -- Buddha
> www.quotator.net
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> DDI-SRG mailing list
> DDI-SRG at icpsr.umich.edu
> http://www.icpsr.umich.edu/mailman/listinfo/ddi-srg


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