When you code something as 1, 2, 3, 4, 5 then it is, at least for me, obvious that it is discrete values. But many people seem to talk about a variable as being “continuous”. I have interpreted such statements as that they mean that the variable is an interval scale variable or even a ratio scale variable.
When we count something then that is obviously a discrete variable. But it is also a ratio scale variable. Four persons are twice as many as two persons.
Think of a visual analogue scale (VAS). For example a ten centimetre long line where you can mark for example satisfaction with a product. Such a variable is continuous. But is it an interval or ratio scale variable?
The economists talk a lot about utility. They think of a utility function as being continuous but they insist that it is only an ordinal variable, not an interval or ratio scale variable.
I think of the discrete versus continuous is one thing and the discussion about ordinal, interval and ratio scale is something completely different.
Or have I missed something here? Do they mean something different?
OP here so i don't really want to get in to whether likert-scale is continuous or not as like i said i'm not a statistician.
However, to my original question, everyone agree that it can be a T-test even though type 1 error will be high? or is there no other test?
For the mann-whitney U test, will i be able to get the same result as I wold with t-test? I need to be able to differentiate based on the demographics (m/f) or (education level) what the preferences of my sample will be.
what about chi-square?
Tweet |