Metadata from Blaise and DDI 3 03 2
Metadata from Blaise and DDI 3. 0/3. 2 Gina Cheung Beth-Ellen Pennell North American DDI Conference April 1 -2, 2014 © 2014 by the Regents of the University of Michigan
Agenda • Blaise Metadata (MQDS) • DDI 3. 0/3. 1 – DDI 3. 2 • Next Step © 2014 by the Regents of the University of Michigan
What is MQDS? • The Michigan Questionnaire Documentation System (MQDS) was designed to extract comprehensive metadata from Blaise survey instruments and render it as an e. Xtended Markup Language (XML) document using the Data Documentation Initiative (DDI) standard. From 2013 NADDI presentation © 2014 by the Regents of the University of Michigan
MQDS Development Timeline Version 1 • 2004 • Blaise. Doc Version 2 • 2007 • . Net Programming • 2009 Version 3 • Database driving and DDI 3. 0 compatible Version 4 • 2011 • Able to handle large data model and DDI 3. 1 compatible Version 5 From 2013 NADDI presentation © 2014 by the Regents of the University of Michigan • 2014? • Able to work with Blaise 5 new structures and DDI 3. 2 compatible
MQDS usage Free to all Blaise community users Questionnaire design Questionnaire documentation Quick data transformation (BDB format to SAS, SPASS or SQL DB format) • Creating interviewers’ training materials • Codebook with universe info and “go-to” logic • • © 2014 by the Regents of the University of Michigan
MQDS Limitations • • Complex Blaise questionnaire roster structure Blaise alien router Paradata documentation Not a tool for Survey life circle documentation yet…. From 2013 NADDI presentation © 2014 by the Regents of the University of Michigan
DDI History • DD 2: – Documents variables in a dataset – Question text from a survey is associated with a single variable – No detailed description of the question or response options – No question flow, only simple forward and backward links • DDI 3. 0/3. 1 – Describes instrument flow – Textual description of conditions for if/then, loops, and dynamic text From 2013 EDDI Colectica’s presentation © 2014 by the Regents of the University of Michigan
Blaise Metadata Show Case • Collaborative Psychiatric Epidemiology Surveys (CPES): to collect data re: distributions, correlates and risk factors of mental disorders as well as explore ethnic disparities, support systems, discrimination, and assimilation – Interviewing extended from 2001 -- 2003 – All probability samples of non-institutionalized adults living in coterminous US – Version of World Health Organization’s Composite International Diagnostic Interview (CIDI) programmed in Blaise • CPES Harmonization Project: – – Harmonize and merge CPES datasets Prepare enhanced tools (interactive codebook) and data for on-line analysis Develop Web site Disseminate for public use © 2014 by the Regents of the University of Michigan
Challenges • Documented 3 instruments with multiple datamodels – Needed to create pseudo instrument for CPES merged dataset – Edited instrument datamodels to improve documentation • Created e. Xtended Markup Language (XML) files with question metadata for 9 XML files – CPES – Individual studies – NLAAS five languages 9 © 2014 by the Regents of the University of Michigan
Complete Process © 2014 by the Regents of the University of Michigan
Interactive Codebook 11 © 2014 by the Regents of the University of Michigan
CPES V 04454 Frequencies 12 © 2014 by the Regents of the University of Michigan
ICPSR Developed Web site • Used XML files and the crosswalk to create interactive codebook • Provided – – keyword search capability downloadable datasets and codebooks online analysis tools links to study documentation, publications, and related sites • Web site maintenance and user support © 2014 by the Regents of the University of Michigan
© 2014 by the Regents of the University of Michigan
New to DDI 3. 2 • The main new things around questions and surveys in DDI 3. 2 are: – Fragments, Groups, Represented variables – Modeling data flows through an Instrument with In. Parameters, Out. Parameters, and Bindings – Question. Grids – Question. Blocks – Scale questions – Ranked questions – Point distribution questions © 2014 by the Regents of the University of Michigan
Fragments DDI 3. 0 <s: Study. Unit id="GITFIELD">. . . <r: Citation>. . . <s: Abstract id="ABS_1">. . . <r: Universe. Reference is. Reference="true">. . . <s: Purpose id="PUR_1">. . . <c: Conceptual. Component is. Maintainable="true" id="CC_1">. . . <d: Data. Collection is. Maintainable="true" id="DC 1" version="1. 0. 0" version. Date="2014 -03 -26">. . . <l: Logical. Product id="LP 1">. . . <l: Category is. Versionable="true" id="cat_17_C 82" missing="false"> <l: Category. Name>Male</l: Category. Nam e> <r: Label xml: lang="en. US">Male</r: Label> </l: Category>. . . © 2014 by the Regents of the University of Michigan DDI 3. 2 <ddi: Fragment> <l: Category> <r: Agency>int. example</r: Agency> <r: ID>Male</r: ID> <r: Version>1</r: Version> <r: Label> <r: Content xml: lang="en. US">Male</r: Content> </r: Label> </l: Category> </ddi: Fragment>
Groups DDI 3. 1 <l: Variable. Group id="var. Group 1"> <r: User. ID type="G">GITFIELD</r: User. ID> <l: Group. Type>section</l: Group. Type> <r: Label xml: lang="English">GITFIELD</r: Label> <l: Variable. Reference> <r: ID>V 1</r: ID> </l: Variable. Reference> <r: ID>V 461</r: ID> </l: Variable. Reference> <r: ID>V 23</r: ID> </l: Variable. Reference> </l: Variable. Group> © 2014 by the Regents of the University of Michigan DDI 3. 2 <l: Variable. Group> <r: Agency>Uo. M</r: Agency><r: ID>VG 1</r: ID> <r: Version>1</r: Version> <l: Variable. Group. Name> <r: String>var. Group 1</r: String> </l: Variable. Group. Name> <r: Variable. Reference> <r: Agency>Uo. M</r: Agency><r: ID>V 1</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Variable</r: Type. Of. Object> </r: Variable. Reference> <r: Agency>Uo. M</r: Agency><r: ID>V 461</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Variable</r: Type. Of. Object> </r: Variable. Reference> <r: Agency>Uo. M</r: Agency><r: ID>V 23</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Variable</r: Type. Of. Object> </r: Variable. Reference> </l: Variable. Group>
Represented Variables <ddi: Fragment> <l: Variable> <r: Agency>int. example</r: Agency> <r: ID>lvar 1</r: ID> <r: Version>1</r: Version> <l: Variable. Name> <r: String>mar. Stat 2000</r: String> </l: Variable. Name> <r: Label> <r: Content>Marital status of respondent in 2000</r: Content> </r: Label> <l: Represented. Variable. Reference> <r: Agency>int. example</r: Agency> <r: ID>rep. Var 1</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Represented. Variable</r: Type. Of. Object> </l: Represented. Variable. Reference> <l: Variable. Representation> <r: Code. Representation/> </l: Variable. Representation> </l: Variable> </ddi: Fragment> © 2014 by the Regents of the University of Michigan
Data flow: In. Parameters, Out. Parameters, and Bindings <ddi: Fragment> <d: If. Then. Else> <r: Agency>int. example</r: Agency> <r: ID>if. Then 1</r: ID> <r: Version>1</r: Version> <r: In. Parameter> <r: Agency>int. example</r: Agency> <r: ID>in. Param 2</r: ID> <r: Version>1</r: Version> <r: Alias>gender. Of. Respondent</r: Alias> <r: Code. Representation> <r: Code. List. Reference> <r: Agency>int. example</r: Agency> <r: ID>code. List. Gender</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Code. List</r: Type. Of. O bject> </r: Code. List. Reference> </r: Code. Representation> </r: In. Parameter> <d: If. Condition> <r: Command> © 2014 by the Regents of the University of Michigan <r: Program. Language>Meta. English</r: Program. Language> <r: Command. Content> gender. Of. Respondent == Female </r: Command. Content> </r: Command> </d: If. Condition> <d: Then. Construct. Reference> <r: Agency>int. example</r: Agency> <r: ID>seq 2</r: ID> <r: Version>1</r: Version> <r: Type. Of. Object>Sequence</r: Type. Of. O bject> </d: Then. Construct. Reference> </d: If. Then. Else> </ddi: Fragment>
Question representations • • • Question. Grids Question. Blocks Scale questions Ranked questions Point distribution questions © 2014 by the Regents of the University of Michigan
Question. Grids © 2014 by the Regents of the University of Michigan
Question. Grids - 3. 1 <d: Question. Item is. Versionable="true" id="Q 29"> <d: Question. Item. Name xml: lang="English">Name</d: Question. Item. Name> <d: Question. Text is. Structure. Required="true" xml: lang="English"> <d: Literal. Text> <l: Variable. Group id="var. Group 5"> <d: Text>What is your name? </d: Text> <r: User. ID </d: Literal. Text> type="G">GIT. HHL[1]</r: User. ID> </d: Question. Text> <d: Code. Domain> <l: Group. Type>array</l: Group. Ty <r: Code. Scheme. Reference> pe> <r: ID/> <r: Label xml: lang="English"/> </r: Code. Scheme. Reference> <r: Universe. Reference> </d: Code. Domain> <r: ID>U 20</r: ID> </d: Question. Item> <l: Variable id="V 29"> <l: Variable. Name>GIT. HHL[1]. Name</l: Variable. Name> <r: Label type="var">Name</r: Label> <l: Question. Reference> <r: ID>Q 29</r: ID> </l: Question. Reference> <l: Representation> <l: Text. Representation max. Length="40"/> </l: Representation> </l: Variable> © 2014 by the Regents of the University of Michigan </r: Universe. Reference> <l: Variable. Reference> <r: ID>V 28</r: ID> </l: Variable. Reference> <r: ID>V 29</r: ID> </l: Variable. Reference> <r: ID>V 31</r: ID> </l: Variable. Reference>
Question. Grids - 3. 2 ddi: Fragment> <d: Question. Grid> <r: Agency>int. example</r: Agency> <r: ID>grid 1</r: ID> <r: Version>1</r: Version> <d: Grid. Dimension rank="1"> <d: Roster base. Code. Value="1" code. Iteration. Value="1" minimum. Required="1" maximum. Allowed="25" > </d: Roster> </d: Grid. Dimension> <d: Structured. Mixed. Grid. Response. Domain> <d: Text. Domain> <r: Label> <r: Content>Name</r: Content> </r: Label> </d: Text. Domain> </d: Grid. Response. Domain> < . . . © 2014 by the Regents of the University of Michigan ddi: Fragment> <d: Question. Item> <r: Agency>int. example</r: Agency> <r: ID>question 1</r: ID> <r: Version>1</r: Version> <d: Question. Item. Name> <r: String>q 1</r: String> </d: Question. Item. Name> <r: Out. Parameter> <r: Agency>int. example</r: Agency> <r: ID>out. Param 1</r: ID> <r: Version>1</r: Version> <r: Alias>first. Name</r: Alias> <r: Text. Representation/> </r: Out. Parameter> <d: Question. Text audience. Language="en-US"> <d: Literal. Text> <d: Text>What is your name? </d: Text> </d: Literal. Text> </d: Question. Text> <d: Text. Domain/> </d: Question. Item> </ddi: Fragment> <
Scale questions - 3. 0 <d: Question. Item id="Q 387"> <d: Question. Item. Name xml: lang="en-US">Keep Up w/ Infl</d: Question. Item. Name> <d: Question. Text is. Structure. Required="true" xml: lang="English"> <d: Literal. Text> <d: Text>On a scale of 0 to 100, what do you think are the chances that your income will keep up with inflation for the next five years? </d: Text> </d: Literal. Text> </d: Question. Text> <d: Code. Domain> <r: Code. Scheme. Reference> <r: ID/> </r: Code. Scheme. Reference> </d: Code. Domain> </d: Question. Item> © 2014 by the Regents of the University of Michigan . . . <l: Variable id="V 387"> <l: Variable. Name>Keep. Up</l: Variable. Name> <r: Label xml: lang="en-US">Scale Intro</r: Label> <r: Label type="var">Keep. Up</r: Label> <l: Question. Reference> <r: ID>Q 387</r: ID> </l: Question. Reference> <l: Representation> <l: Numeric. Representation type="Integer"> <r: Number. Range> <r: Low>0</r: Low> <r: High>100</r: High> </r: Number. Range> </l: Numeric. Representation> </l: Variable>
Scale questions - 3. 2 <ddi: Fragment> <d: Scale. Domain> <d: Question. Item> <r: Scale. Dimension> <r: Agency>int. example</r: Agency> <r: ID>Q 1</r: ID> <r: Number. Range> <r: Version>1</r: Version> <r: Low>0</r: Low> <d: Question. Item. Name> <r: High>100</r: High> <r: String>Keep. Up</r: String> </r: Number. Range> </d: Question. Item. Name> <r: Value. Increment start. Value="0" <d: Question. Text audience. Language="en-US"> end. Value="100" increment="1"/> <d: Literal. Text> </r: Scale. Dimension> <d: Text>On a scale of 0 to 100, what do you think are the chances that your income will keep up with inflation for the <r: Label> next five years? </d: Text> <r: Content>Keep Up w/ Infl</r: Content> </d: Literal. Text> </r: Label> </d: Question. Text> © 2014 by the Regents of the University of Michigan </d: Scale. Domain> </d: Question. Item>
Ranked questions <ddi: Fragment> <d: Question. Grid> <r: Agency>int. example</r: Agency> <r: ID>grid 1</r: ID> <r: Version>1</r: Version> <d: Grid. Dimension rank="1"> <d: Roster base. Code. Value="1" code. Iteration. Value="1" minimum. Required="0" maximum. Allowed="100" > </d: Roster> </d: Grid. Dimension> <d: Structured. Mixed. Grid. Response. Domain> <d: Text. Domain> <r: Label> <r: Content>Name</r: Content> </r: Label> </d: Text. Domain> </d: Grid. Response. Domain> © 2014 by the Regents of the University of Michigan <d: Grid. Response. Domain> <d: Ranking. Domain> <r: Ranking. Range> <r: Minimum. Value included="true">1</r: Minimum. Value> <r: Maximum. Value included="true">5</r: Maximum. Value> </r: Ranking. Range> </d: Ranking. Domain> </d: Grid. Response. Domain> <d: Distribution. Domain> <r: Distribution. Value>100</r: Distribution. Val ue> </d: Distribution. Domain> </d: Grid. Response. Domain> </d: Structured. Mixed. Grid. Response. Domain> </d: Question. Grid> </ddi: Fragment>
DDI 3. 2 is much more powerful to Document the full survey lifecycle in a standard manner © 2014 by the Regents of the University of Michigan
Next Step • • DDI 3. 2 MQDS 5? Blaise 5? Colectica 5? Thank YOU! © 2014 by the Regents of the University of Michigan
- Slides: 28