Rescaling a 7-Point Likert Scale

[Serena Menny]

Hello everyone,

I have conducted a survey where all responses are based on a 7-point Likert scale. To test the robustness of my findings, I am considering rescaling the responses to a 4-point Likert scale, eliminating the neutral option.

Does this approach make sense for my analysis?

If so, could anyone provide guidance or Stata code for performing this rescaling?

Thank you!

[Erik Ruzek]

Serena,

My immediate reaction is that this is not a good idea, but you need to say much more about how you think throwing out valid responses to a survey item is a test of the robustness of your findings. What kind of analytical model are you running?

In working with item response data such as yours, and doing psychometric analyses, it can sometimes be necessary to drop a response option in an analysis when you are testing whether the measure is valid among different groups of respondents. For example, you might find that people in one group never utilize the neutral option. In that case, you would drop the neutral option only for those in that group when conducting the factor model. But that is a very special case.