Dealing with change in Likert scale

Someone on our team thought it was a good idea to change the Likert scale for a set of questions (interview data) from a 7-point to a 5-point scale. My first thought was to just standardize all this data and analyze from there, except that it would be nice to be able to compare back to the (now) 5-point scale to say things like, control group had a mean of 2.3, which relates to a response of 'rarely', and experimental group had a mean of 4.2, which relates to a reponse of 'very often' on XXscale.
I know in Descrpitives you can ask for the standardized scale (z-score), but not sure how I would go about converting back to my 5-point scale. Any ideas?:confused:


TS Contributor

Well, the thing is that you cannot find the mean in a Likert scale data, what you can do is find the mode (most frequent value).


TS Contributor
Well, the thing is that you cannot find the mean in a Likert scale data, what you can do is find the mode (most frequent value).
Actually, if the underlying concept being measured can be considered continuous, you can compute descriptive stats such as the mean and not be worried too much about the most cases a Likert scale can be treated as an interval-level scale without concern for drawing incorrect conclusions....

...however, trying to compare two different Likert scales - that's an entirely different story, because of the way different people interpret the phrases next to the scale numbers, how they "use" the scales when answering the question (the bigger the scale, the less of it gets used), etc., so I would completely avoid trying to compare a 5-point to a 7-point on an apples-to-apples basis, even after "normalizing" them.
we create scales from Likert scale data and use the mean all the time in behavioral research - but understand that for interpretation purposes, mode is better. The idea is that we are looking at a continuous scale.

Probably not the best thing to try to compare these scales, I know, but am not in charge or I would have vetoed the change in scale to begin with. (we also had an issue with someone creating a scanable form for us and they missed a value - turning a 6 point scale into a 5 point scale. Busy, busy and last minute getting ready so no one noticed until all surveys returned - really wary of messing with that....). Sometimes I wonder why I ever decided that doing research with people was better than with animals! Animals don't complain when you have to redo a test - people will often outright refuse.