I'd think using plots to test the assumptions of multiple linear regression requires some experience anyway, but in this case all my independent variables, and my dependent variable, are 5-level Likert rather than continuous variables.

I understand applying non-ordinal regression to this kind of data is somewhat controversial, but is often done.

The plot consists of 5 linear arrays of points at 45 degrees. Is that interpretable at all?

If not, what should be substituted?

  • 1
    $\begingroup$ Yes,thanks and sorry. That post IS a duplicate, covering the same ground, and as suspected the pattern IS apparently due to non-continuous data, $\endgroup$
    – Ed Lithgow
    Jul 25, 2019 at 6:30
  • $\begingroup$ Hi, here is even a more upvoted thread about the same "straight lines" phenomenon stats.stackexchange.com/q/33165/3277 $\endgroup$
    – ttnphns
    Jul 25, 2019 at 7:10
  • $\begingroup$ Visit this too stats.stackexchange.com/q/287669/3277 $\endgroup$
    – ttnphns
    Jul 25, 2019 at 7:13
  • $\begingroup$ Ed, if you think it is a duplicate and your question isn't going to be informative, then you could ask a moderator to close or delete it. $\endgroup$ Jul 25, 2019 at 19:22


Browse other questions tagged or ask your own question.