I am researching EI and Humility. My EI assessment used a 7 point Likert scale and my humility assessment a 5 point Likert scale. Do I need to convert the 5 point to a 7 point scale to be able to run regression analysis? Clearly the means for the 5 point scale would be lower than the mean for the 7 point. Am I trying to compare apples to oranges if I don't convert the scales to like properties?
Thanks for responding!
Dear Jill,
Safe question: Yes. If you want to meaningfully compare both, there needs to be some kind of relativization. But usually in a regression most programs return standardized regression weights... (which are the weights when the variables are standardized beforehand). If the regression does not give you this automatically, then standardize the variables beforehand (to mean 0 and SD=1).
Keep in mind, in most cases unstandardized weights can nonetheless help interpreting the single effect size of a predictor in a natural way (in the sense of the original scale units).
Hope this helps
Best, René
Jill Kähli, Rene has given you some good advice up above. In essence, however, I think you needn't worry. Certainly, don't go to the effort of trying to convert either of the scales so that they match each other in terms of the maximum obtainable score or range of scores.
In essence, take a deep breath, have a sigh of relief, and just keep going.
All the best with your research.
There is an important difference between analyzing individual, Likert-scored items versus scales that are composed of multiple items. If you have multiple item scales, you can calculate average scores, which would then be on 1-5 and 1-7 metrics, which can be easier to interpret.
Also, as René Schlegelmilch points out, you could rely on standardized regression coefficients. Or, if you want to preserve the metric of the independent variables, you could standardize the scores for the scales averages by first calculating the average across the items (as above), and then dividing by the standard deviation of those sums.
Kindly check https://www.theanalysisfactor.com/likert-scale-items-as-predictor-variables-in-regression/
Also check https://www.ibm.com/support/pages/transforming-different-likert-scales-common-scale
Most multi-variable survey research uses differently scaled measures for regression analysis, etc., no problem.
It is preferable to measure different variables with different scale points to avoid common method bias. However, while interpreting regression coefficients, you should report standardized regression coefficients if you measure different variables using different scales.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Social Media