Capture Output of a CAI System

Ask an Expert for Help

Computer Aided Interviewing is increasingly the preferred method of data collection in larger studies, through dedicated software such as Blaise [http://blaise.com/], IBM Data Collection [http://www-01.ibm.com/software/uk/analytics/spss/products/data-collection/] or CASES [http://cases.berkeley.edu/], and increasingly through web-based software such as REDCap [https://catalyst.harvard.edu/services/redcap/] or LimeSurvey [https://www.limesurvey.org/en/]. 

Documenting the process of data collection is essential. As well as capturing what was implemented in the survey collection (as opposed to what was specified), it is important in provenancing the output variables and reusing this. Without the need for manual entry this is obviously also a great time saver. Unfortunately, none of the software packages support DDI-L natively, but much can be extracted from them that can be used to capture the majority of the questionnaire used.

The way in which CAI software is programmed can make a big difference to what is retrievable at the end of a survey. For instance, for convenience and display purposes, questions and interviewer instructions are often concatenated together. Splitting these out as separate items when programming will make extracting these metadata elements more straightforward.

A further structural problem with most CAI software is that it separates out the flow controls of the questionnaire from the questions and instructions, so software that converts to DDI-L where these are unified into a single model will need to combine them.

Tools and Examples

Blaise (using the Michigan Questionnaire Documentation System (MQDS) plugin). A version that supports Blaise 5 and DDI-L 3.2 is currently in development, although earlier versions of MDQS export out DDI-L 3.1, which can be transformed to DDI-L 3.2 using tools such as SledgeHammer and Colectica Designer. Colectica Designer also has a plugin that allows extraction of DDI-L 3.2 from a Blaise (bim) source file.

IBM Data Collection has its own proprietary XML format which contains most of the information you would want to extract to DDI-L. Colectica Designer also has a plugin that allows extraction of DDI-L 3.2 from a IBM Data Collection (mdd) source file.

LimeSurvey has its own internal XML format. This is not as rich as the most current version of DDI-L, but they have produced tools that will convert to DDI at [https://surveys.acspri.org.au/quexmltools/]. Colectica Designer also has a plugin that allows extraction of DDI-L 3.2 from a LimeSurvey (queXML) source file.

REDCap has arguably a less sophisticated export format than the previously mentioned CAI software, but again contains much which can be repurposed to DDI-L. There is an R package available from  [http://kuscholarworks.ku.edu/handle/1808/11047] which converts a REDCap (csv) source file. Colectica Designer also has a plugin that allows extraction of DDI-L 3.2 from a REDCap (csv) source file.

CASES is currently being redeveloped, DDI files can be imported into CASES. Colectica Designer has a plugin that allows extraction of DDI-L from a CASES (.q;.m.FILES) source file.