Implications of eliminating the middle option of a 5-point Likert

#1
Hi all,

Assuming I have data from a multi-item questionnaire, 5-point Likert.
Raw scores have been converted into STEN and tabulated into norms.
I now remove the middle option of the 5-point scale and convert it to a 4-point scale.
Can new data from this 4-point scale be added to the existing norms, which were created with 5-point scale data?

I felt that eliminating one option changes the spread of variance and thus the new data should not be used.

In this case, does anyone know of a technique that can somehow "pro-rate" the new data such that it does not face this "spread-of-variance" violation?
 

JohnM

TS Contributor
#2
It is not recommended practice to change scales after the fact in order to "fit" them to another data set or larger database. Different people interpret different scales differently......
 
#3
Thanks John.

To give further background, i am looking at converting the 5-pt scale of a 20 question survey down to a 4-pt scale.

I ran reliability checks (cronbach's) and found a higher alpha when only including cases that had no responses on the middle option versus including the entire sample.

Can this serve as some form of justification to the removal of the mid-point?