My research involved a questionnaire with open-ended questions, multi-answer multiple choice (select all that apply; entered as "1" for circled answers, "0" for non-circled), and likert scale questions. I want to examine whether there are differences in responses to the survey (especially the likert questions) based on a variety of independent variables. Independent variables are things such as: age, ward number, intervention group (3 colour-coded groups - this was a randomized clinical trial), treating physician/dietitian (entered into SPSS as string variables), presence/absence of dysphagia (entered as binary data 1,2)...and many others.
I would like to treat the likert data as ordinal data rather than interval. This is my first time conducting a research study and would appreciate pointers as to how to even begin the analysis. My limited understanding is that non-parametric tests are more suitable for my scenario.
Does your likert scale have some kind of meaning between successive measurements? (i.e. is there some significant way to interpret the difference between say a 2 and 3 on some kind of scale), or is there no way to do such an interpretation even though there is a ranking attributed to the variable itself?
I'm not sure there is a method of saying for certain that the difference between 2-3 is same as the difference between 4-5, which is why I've chosen to treat it as ordinal. Was that what you were referring to?
It doesn't have to be fixed, it can be some kind of relative interval: the main thing is that there must be some kind of interval property between two successive outcomes. If there is not such a property, but the data can still have a "ranking", then the data is ordinal.
For example you could have an exponential scale in your data and when you take the logarithm of this data, you get an interval difference that looks like the counting numbers so this example demonstrates that the interval difference may not be linear, but it can be made linear if need be.