What I want to tell you is a little bit about some research we did in Europe with respect to questionnaire documentation and—to be more specific—questionnaire instrument documentation. The background of this is the same as was already discussed this morning.
We changed, in the Netherlands and in Europe, to computer-assisted interviewing—and in particular to computer-assisted personal interviewing with laptops—in the 80s of the previous century. And we had, from [that] moment, a growing complexity of questionnaires [and] growing size of questionnaires. We have seen, a famous example, instruments with 10,000s of questions. And indeed the old paper questionnaire was more-or-less self-documenting. But these large, complex questionnaires have become more and more of a problem from the point of view of documentation.
So the basic question is: how can you make, for a complex, large questionnaire instrument, a human-readable documentation? And we wanted to have a solution for that and started a project to develop a prototype for that, and that became TADEQ. So we want to generate, to create a software tool for questionnaire documentation. And, when we thought about it, we realized that once you have a tool for questionnaire documentation it’s also a very useful tool in the process of developing a questionnaire. Because more or less the same aspects with respect to documentation, and getting insights into complex structure of the questionnaire, are involved in both the final documentation and in the process of developing it.
What we tried to do—as Tom also explained—was that we started in a later phase of thinking on this issue, and tried to have a more-or-less object-oriented approach in developing a questionnaire documentation tool. We see a questionnaire instrument as a collection of objects—all kinds of objects you can have in a questionnaire. Questions, checks, route instruction, computations—whatever you can think of. And all these objects are part of a kind of questionnaire execution tree.
So here I’ve named a number of these types of objects. You have questions—the various types of questions [such as] open, floats, numeric, etc. Route instructions, which can either be GOTO-oriented instructions as in CASES or IF-THEN-ELSE structures like in Blaise. We have checks in the questionnaire structure; these were not mentioned too much this morning, I think, but these are a valuable, extra advantage as compared to a paper questionnaire. You are able to detect inconsistencies while you are carrying out an interview, and are also able to correct incomplete answers in the course of an interview. If you have to do that later on, in the office, it’s almost impossible to do that. So we feel that, with respect to checks, a questionnaire instrument can add to the quality of the final collected data. You have computations, you have loops. And