Trials and Tribulations of creating DDI Codebooks at the University of Guelph

1
Trials and Tribulations of creating DDI Codebooks
at the University of Guelph

• A.Michelle Edwards and Carol Perry,
• Data Resource Centre,
• University of Guelph
• Guelph, Ontario

2
Current Search Function
3
Search Results
4
Current Documentation
5
Identifying Variables
6
Rationale for Change

• 522 datasets to date.
• No comprehensive metadata search function.
• No current variable search within dataset.
• Limits researchers autonomy.

7
XML tags

• Started with approx. 30 or so tags
• As of June 5, 2002
• 101 tags
• 59 are filled
• Information contained inside tags

8
Codebook Templates

• Used Maddie to develop initial template.
• Edited the template to add tags as required.
• Filled in fields common to all codebooks.

9
Codebook Templates

• Statistics Canada data
• ICPSR data
• B2020 data format

10
Statistics Canada Codebook
11
Differences between Codebook Templates

• Authoring entity
• Distributor (DLI vs. ICPSR)
• Licenses
• Other material ICPSR abstract link
• B2020
• No direct link to database
• No variables

12
How do we move our information from an HTML
readme file to an XML file???
13
Readme to XML

• Document Description
• Study Description
• Data Files Description

14
Readme to XML

• Currently copy and paste information from the
  Readme (html) file into the XML Codebook.
• Script extracts metadata from html and places
  into XML.
• Same amount of time.

15
Variable Information
16
Variable Information

• Sources of Variable information
• Variable names, labels, and position from the SAS
  program.
• Frequencies for each variable value from SAS
  output.

17
Variable Information

• Sources of Variable information
• Literal questions from questionnaires if
  available.

18
Variable Information

• Script
• Looks into the SAS program pulls out the
  variable names, labels and positions.
• Looks into a SAS output file for frequencies and
  variable value labels.

19
Variable Information

• Script
• If questionnaire is available seeks out
  questions and matches with variables.

20
Variable Information

• Problems with Script
• SAS programs must be consistent in their format.
• SAS output and questionnaires matching
  variables.

21
SAS to XML

• SAS 8.2 - XML engine and ODS XML.
• Can create XML SAS output.
• Variable names, labels, value labels, and
  frequencies.
• Variable positions with the input statement and
  Proc Print ? XML.

22
SAS to XML Frequency Output
23
SAS to XMLProc print output
24
SAS to XML
25
SAS to XML

• Advantages
• SAS programs do not need to be consistent.
• Use one program from start to finish SAS.
• Still in development.

26
XML to Viewable Document

• Saxon to render our XML documents to HTML using
  XSL Stylesheets.
• XSL pull out info from XML document and display
  with HTML tags.

27
XSL Templates

• Set for each
• Statistics Canada
• ICPSR
• B2020
• Initial templates from University of Virginia
  samples.

28
XSL Templates

• Abstract
• Study Info
• Methodology File Dimensions
• Questions
• Variables Frequencies
• Other Documents

29
XSL Stylesheets
30
Search

• Uses SAS IntrNet to call and run the UNIX SGREP
  search.
• Creates an XML file with results.
• Calls Saxon to render the file with the Variable
  XSL Stylesheet.

31
Final Product

• Frames to put it all together.
• Links to each component (abstract, etc.).
• Returns the rendered HTML on the fly.

32
Final Product
33
Final Product

• Sun Exposure Survey 1996
• http//tdr.uoguelph.ca/DATA/WWWDOCS/XML/SES2/ses96
  cbk.html

34
Finished Product

• 522 datasets to date.
• 35 Completed DDI-compliant codebooks.
• Fall completion ???

35
Final Product
36
Final Product
37
Final Product
38
Final Product
39
Final Product
40
Final Product
41
Final Product