I just conducted the power analysis for a lab study I am hoping to run soon. The main idea is comparing the difference in means across three conditions with a balanced one-way ANOVA (see image below).

Basically, I am amazed at how underpowered I am with moderate effect sizes. Unless I set large effect sizes (f=0.4), I am severely underpowered.

To help overcome these power issues, I'm wondering if bootstrapping may be an appropriate method of resampling? While I recognize the lab study participants are not nationally representative of the broader public, I'm more interested in the psychological phenomenon that occurs.

What do you all think--would bootstrapping be one method of overcoming this potential issue with power? Open to any and all suggestions!


enter image description here

  • 2
    $\begingroup$ No, bootstrapping will not in general improve power. Not answering formally as there have been previous discussions here, here, and here. If you really cannot collect more data, I can recommend the book Small Sample Size Solutions (2020), which notably does not include a chapter on the bootstrap (because it's not a good small sample technique). $\endgroup$
    – awhug
    Oct 12, 2022 at 13:12
  • $\begingroup$ Thank you! I tried looking for an answer in other places but couldn't find one. Much appreciated. $\endgroup$
    – JimStir
    Oct 12, 2022 at 19:23


Your Answer

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