If the data for the continuous variable is not normally distributed (shapiro-wilk sig = .000) can you still use a point-biserial correlation??

I have run the point biserial correlation to see and it is significant (p=.006) - so does this mean that it does not matter if the data is normally distributed??

If not what test would I use instead??

  • $\begingroup$ Point-biserial r is usual r used to correlate one variable dichotomous the other continuous. As with r, classic asymptotic significance test would assume normal distribution for the continuous counterpart. However, the test is robust to not strong violations of normality. $\endgroup$
    – ttnphns
    Nov 9, 2018 at 20:20
  • $\begingroup$ Hi, thank you for your reply! How do you find out if the violations of normality are strong or not? How do you find out if it is still ok to use point-biserial correlation if your data is not normally distributed? Do you have any links/could you explain this further?? Thank you for your help! $\endgroup$
    – Millie
    Nov 9, 2018 at 21:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.