I have composited Likert scales from many items for some employee motivation variables.

what is the most suitable way to build the scale from discrete Likert items? is it using the mean or the median?

I need to study the correlation between motivation and burnout Likert scales or variables. what is the most suitable correlation coefficient for this case? Pearson or Spearman?



1 Answer 1


To create a scale from several Likert items, people typically sum or average the items. This introduces a theoretical problem, since it involves assuming that the categories on the Likert items are equally spaced. But that's something most people can live with. Using a mean is useful if there are some respondents with some missing items. Using the median would be theoretically more responsible, but would probably result in less diversity in the scale values.

From there, you can use Spearman correlation, unless the data are bivariate normal, in which case Pearson is also appropriate.

  • $\begingroup$ Thank you. Actually i applied Spearman and Pearson correlations on my composite variables. the correlation coefficients were near to each other but Spearman r had more significant results at p<0.01 than Sprearman's one. Could it be because after taking the mean to composite Likert items the data is not ordinal any more and it is more continuous? $\endgroup$
    – Wael Ashi
    Aug 26, 2017 at 9:14
  • $\begingroup$ The difference in result between Pearson and Spearman can be related to the relationship between the two variables. Pearson is most appropriate when the relationship between the two variables is linear. Spearman can identify relationships that are monotonic but not linear. The relationship can be determined simply plotting one variable vs the other in a scatter plot. $\endgroup$ Aug 26, 2017 at 10:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.