I was just wondering, for those with experience in the area, how do you best deal with "dont knows" in a survey. I have a Likert scale ranging from 1-7 across many questions, but the modal answer for most questions is 0 = dont know/care. I was thinking of dividing the scale into four components...Dont Know, Satisified, Neutral and Dissatisfied and treat this as a categorical variable, however there are many questions and only 400 respondents. Presumably I could test each variable at a univariate level and discard the "worst" and then work with whats left, but I'd appreciate some advice from anyone that works in this area.
Kieran
|