Is it more reasonable to remove "IDK" from a survey in SPSS for rudimentary analyses?

Question

I attached a graph showing the typical array of responses to a survey measured ordinally in SPSS. While generally, the answers are quantifiable/compared as 1–5 corresponding to answers ranging from "Not Satisfied" to "Very Satisfied," many questions have a 6 (equating to "IDK"), although it's usually a minority response. I worry that for my purposes it's just noise and distracting from the 'weight' of other, more certain answers.

So, does it make statistical sense to measure opinions this way if I'm looking for ordinal certainty on a 0-6 scale? I'm wondering now if it's best to recode anything, change the Measure, or remove all 6s from the data totally.

Thanks for checking this post out!

Answer 1

A Likert scale assumes that response categories have a rank order. The survey question is apparently measuring the satisfaction of Monk Seal Resting and Sea Turtle Nesting, and you can't really rank "I don't know" on a satisfaction scale, where, for instance, 1 might represent "Not Satisfied" and 5 might represent "Very Satisfied".

If the purpose of your analysis is to measure statistical differences or levels of agreement between groups, then you probably want to remove "blank" and "I don't know". If you are creating a visualization to show the survey responses, and it would be helpful for your audience to know that a number of people chose "I don't know", and a number of people chose not to respond to the question, then it might make sense to keep those responses in your analysis. I would recommend reordering them so that they appear together at the beginning or end of your graph, instead of sandwiching the true Likert scale responses.