I am going to conduct a survey in order to collect information for two variables and then calculate the correlation between them.

If we assume that $alpha$ = 0.05, $power=0.80$. Can someone show me a step by step method of how a sample of 85 was determined with a medium effect size? And is there a procedure in SAS I can use?



1 Answer 1


In SAS you can use PROC POWER. Something like

proc power;
   onecorr dist=fisherz
      nullcorr = 0.00
      corr = 0.10 to .9 by .1
      ntotal = .
      power = .8;

seems to do what you want. Definitions of 'medium' can, of course, vary; many people use the definition suggested by Jacob Cohen.

  • $\begingroup$ I ran the code, and here is a line from output: r= 0.3, alpha=0.0500, power=0.800, n=84. The sample size is 84, but I'm a bit confused about r=0.3, I don't know what's the value of r, but here it's specified to be 0.3. Can you please explain this? thanks $\endgroup$
    – user9292
    Sep 29, 2012 at 22:00
  • $\begingroup$ r is the correlation coefficient. That is, if you are testing if a correlation is different from 0 - nullcorr = 0.00 - with power = .8 and alpha = .05 (default) then the minimum sample size is 84. $\endgroup$
    – Peter Flom
    Sep 29, 2012 at 22:07

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.