I have two variables, and I can calculate e.g. the Pearson correlation between them, but I would like to know something analogous to what a t-test would give me (i.e. some notion of how significant the correlation is).

Does such a thing exist?

  • 1
    $\begingroup$ Use the Fisher Transformation or a t-test based on a scaled tangent tranformation: en.wikipedia.org/wiki/… $\endgroup$ – whuber Jan 17 '11 at 19:11

Yes, you can get a $p$-value for testing the null hypothesis that the Pearson correlation is zero. See http://en.wikipedia.org/wiki/Pearson%27s_correlation#Inference.

  • 11
    $\begingroup$ For anyone else interested: this can be done in R with cor.test() $\endgroup$ – Xodarap Jan 17 '11 at 21:58

As an alternative, this beautiful paper (a bit technical/mathsy, so beware) allows you to test for any correlation (not just $\rho=0$) using a Non-informative Bayesian Decision Theoretic approach.


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.