Converting from a 7- to a 5-point scale

Rich

New Member
#1
I would appreciate the community’s assistance. I work in a unit that evaluates many training activities. In the past we typically used a five-point Likert scale in our assessments but found that responses provided little variance and were skewed, with most responses being the two highest, positive values. In other words, our participants like the training they receive, believe it is relevant, effective, etc.

To increase the variation in responses, we now use a seven-point Likert scale. To compare this year’s responses with previous years (and the five-point scale), we often convert the responses from the seven-point scale to a five-point scale in the following manner:

1 to 1; 2 and 3 to 2; 4 to 3; 5 and 6 to 4; and 7 to 5

Is this transformation valid? Do we violate prevailing statistical norms? Are there more appropriate ways to rescale the responses so that we can legitimately make comparisons over time?

Thank you in advance for your counsel.
 

JohnM

TS Contributor
#2
Rich,

I don't believe you should do such a conversion - you may be able to convince yourselves that it looks logical from a visual standpoint, but it's quite arbitrary.

More importantly, the respondents will interact with and perceive different scales differently.

If you were not happy with the lack of variance and skewness of the 5-point scale, then I would question why you are bothering to do the comparison at all.

Just go with the 7-point and move forward....

John