Hi all,
Just recently started working with someone who has already written up a set of questions, gathered responses for a “test run”, and now collected a non-test run sampling of responses. My issue is that survey/questionnaire validation is a new area for me, but I have some vague notion of the required process before something meaningful such as inference or even descriptive statistics should be interpreted for the survey responses.
Can anyone provide some general advice (or link to tutorials) and a framework for validating the survey prior to analyzing and interpreting any kind of statistic for the items?
Thanks!
Questionnaire design is a bit of a black art, and my standard response is to say “don’t – find one that’s been validated”.
That said, there are many guidelines in writing questionnaires, and I’d recommend the reading list in the CDC’s brief guide.
The QAS-99 checklist is a very good template for shaking down potential problems with design.
You can also learn a lot from using a “think-aloud” protocol, which can reveal questions that are going to be misinterpreted or cause difficulty in selecting responses.
2 Likes
I wrote a wee little book that doesn’t present the in-depth discussions around research methodology but does summarize best practices. Here is a PDF of a proof…https://www.dropbox.com/preview/Dropbox/Write%20Better%20Surveys.%20Period-shared.pdf?role=personal
1 Like
I’m sorry but the link doesn’t seem to work. Did you create it with Dropobox’s “share” feature?
Unfortunately, I’m unable to access the file for some reason.
This site does not allow posts from anonymous users. To keep your post from being remove please edit your profile to specify your real first and last names.
Making it explicit for other people: It worked!
1 Like