Dictionary parsing: special values in value sets and more

Discussions about tools to complement CSPro data processing
Forum rules
New release: CSPro 8.0
Post Reply
sradyakin
Posts: 1
Joined: May 20th, 2013, 2:39 pm

Dictionary parsing: special values in value sets and more

Post by sradyakin »

To whom it may concern.

I have a couple of questions regarding interpretation of special values in the CSPro dictionaries for an external tool to be able to read CSPro dictionary without CSPro. Any help or advice is greatly appreciated.

1) CSPro declares that there exists a special value 'DEFAULT'. Is it possible to define it somewhere in the dictionary? Such as if the value in the data file is '-' meaning 'DEFAULT' then value 1960 is exported into the export dataset? Or is it up to the user and the value '-' must be exported? (not convenient in numeric data)

2) Given the following dictionary entry for value set, what is the meaning of 99? Is it a special value??

[ValueSet]
Label=Numeric question text
Name=ABCDEFG
Value=0:60
Value=99;MISSING
Name=AG_MISSING,Special

In my understanding the Value=99 must be followed by a user-defined text label, while the next following Name= statement may declare the special identifier (one of DEFAULT, MISSING, or NOTAPPL) followed by a comma-Special marker. However the above extract does not conform to this specification and CSPro 5.0 does not show this value (99) as anything special. How is it possible?

3) The role of the blank label is not clear to me. What is the purpose of defining just one empty value label?
Consider e.g the following case (here whole and only value set defined for an item)

[ValueSet]
Label=Numeric question text
Name=ABCDFEFG
Value=1:50

4) How does CSPro data export facility decide on the choice of a value set, when multiple value sets are defined for the same item? I do not see this option in the interface. Neither of my guessed strategies seems to be happening: neither first, nor last, nor common part of all value sets. (I have a dictionary file with ~18 valuesets defined for some variables, so it is not easy to anticipate anything the CSPro will produce in the export).

5) How are labels assigned to non-integer values treated when exporting data to e.g. Stata, which does not support such labelling?

6) How are labels to interval values are treated when exporting data to e.g. Stata, which does not support such labelling? (I currently round interval ends inwards, which may result in empty intervals, such as in case of [0.2;08] and hence the label is lost :( )

7) If any attempt is done in 6, how do intervals overlaps are treated? (or e.g. interval within interval cases)?

Thank you,
Sergiy
Post Reply