I have a dataset obtained from a survey, where all questions are likert-type, but with different scales (1-5 or 1-2). Is it possible to run k means clustering if I rescale all to 1-6?

Additional questions: Would it make more sense to run PCA first and apply clustering to the PCs?

  • $\begingroup$ The thread I marked as a duplicate of your question seems to answer it very well. Your question boils to "do I need to scale the data for clustering" and in many cases, the answer would be "yes", as described in the thread. $\endgroup$
    – Tim
    Jul 7, 2021 at 12:24
  • $\begingroup$ Yes, I understand that rescaling is important. What I cannot figure out is if it makes sense for likert data. Maybe using PCA first would be more appropriate? $\endgroup$
    – Irene
    Jul 7, 2021 at 13:19
  • $\begingroup$ Yes, you can scale such data. You can even scale binary values. Scaling is about having the same properties like min & max or mean & sd for numerical reasons. $\endgroup$
    – Tim
    Jul 7, 2021 at 13:22


Browse other questions tagged or ask your own question.