I did not get many responses over how to compute $p$ values given Spearman $\rho$ and Pearson $r$ so I will try something else:

I see that when you are using the Spearman $\rho$ value, you can use this to determine critical values depending on $n$:


However I also see this for Pearson's $r$:


How would I compute these directly (given arbitrary significance level, degrees of freedom, $r$, etc) instead of relying on a pre-computed table?

I see absolutely nothing online illustrating how this is done. I am comparing my results against these calculators (which work for small sample sizes): www.socscistatistics.com/tests/

  • $\begingroup$ For Pearson's, calculate the standard error and then get the t-statistic. That can be used to get the p. For Spearman, I'm not sure. (I will try to write an answer later, if no one else does.) $\endgroup$ – Jeremy Miles May 5 '15 at 18:48
  • $\begingroup$ @JeremyMiles Are you referring to SE = 1/sqrt(n-3) and t = r*sqrt((n-2)/(1-r^2)) ? $\endgroup$ – user62753 May 5 '15 at 18:50

The standard reference for Spearman's distribution is Best, D.J. and D.E. Roberts (1975) "Algorithm AS 89: The Upper Tail Probabilities of Spearman's rho", Applied Statistics, 24:377-379. You can read in JSTOR for free.

  • $\begingroup$ I read through this reference and it does not seem to illustrate anything not already mentioned, for instance t = r*sqrt((n-2)/(1-r^2)). I don't know how to use this to compute a p value or a critical value for r. $\endgroup$ – user62753 May 5 '15 at 19:02

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.