SPSS and questionnaire funneling?

Question

Hi, I'm using SPSS to process some data for my research. The questionnaire I used employed funneling: depending on the answer to a question it would direct the respondent to different parts.

My question is how I code the data for SPSS. Inevitably this approach creates missing values to some questions. Do I code these as missing values? Do I create different groups of respondents and study the results independently? Or do I assign a different code for "skipped" items that derived from the redirection?

Any help will be much appreciated :)

Answer 1

Well, there are a number of ways to deal with missing data.  First, you can "impute" data based on how people responded to other answers or based on the mean response.  This generally isn't seen as a very good option.  The overall most popular and most reliable (according to Paul Allison, sociology stats wiz) is to use listwise deletion.  It basically just deletes the entirety of the case if even one answer is missing or not known.  The assumption is that the people who have missed or not answered a question will be randomly distributed throughout the sample.

Hope that helps.

Answer 2

I think the standard practice in this kind of situation is to have (at least) two separate missing value codes, like "NA" for questions  that are skipped via your funneling procedure and just "missing" for questions where they refused to answer or whatever.

There's no disadvantage to having everything in a single data file, because it's fairly easy to filter cases to temporarily break them down into groups if any situation were to come up where you really did want to look exclusively at respondents with valid answers to particular questions.

You can see at the link below that the General Social Survey uses this kind of coding scheme.