If we don't care about the probability of finding a real effect if there is one, AKA a statistically significant finding, why care about statistical power and power calculations?

I mean there are obviously benefits to having larger samples such as decreasing our standard errors and increasing precision etc, but would power analyses matter if we don't care about the p-values? What if we were to just focus on effect sizes and other descriptive statistics?

  • 3
    $\begingroup$ Power is the probability of rejecting the null hypothesis given some particular alternative holds. If you're not testing, you're can't be rejecting hypotheses, so what do you then mean by power at all? Can you clarify what you're asking (and you should indicate where your premise comes from -- why you think anyone does care. Maybe they do, but what leads you to think they do?) $\endgroup$ – Glen_b Nov 27 '17 at 9:00
  • $\begingroup$ AKA a Statisticcally significant finding - please elaborate. just to understand what is meant by p- value ? $\endgroup$ – Subhash C. Davar Nov 27 '17 at 11:03

p-value continues to be debated. see one of links below:

Which of the (two) conflicting p-values should I use when estimating lmer using R

I agree with your assertion in para I of your question. power-test does not serve any purpose if you are interested in checking whether the effect-size is significant or insignificant. The results of power- analysis - p-statistic have a different end use. p-value itself is not dangerous but it is so often misinterpreted by users that its utility has been challenged. Your second paragraph seems to convey a different purpose of power analysis/power different from what it is meant for. welcome to raise further queries in the present context.


Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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