Read on Twitter
I can tell there is a real need for more scientific training regarding power analyses, particularly the type that are used to a priori determine sample size. There’s a lot to cover, but here are some of my favorite resources to start. A thread. /1
@theanalysisfactor has some great workshops about power. You have to pay for them, but IMO they are well worth the money. https://www.theanalysisfactor.com/june-2018-fundamentals-sample-size-calculations/ and https://www.theanalysisfactor.com/august-2018-power-analysis-and-sample-size-determination-using-simulation/ /2
@theanalysisfactor also has some great (free) blog posts about power, effect sizes, and sample size. I find their posts to be super clear and informative in general (and they cover lots of topics besides power too!). You can see a list here. https://www.theanalysisfactor.com/resources/by-topic/effect-size-statistics-power-and-sample-size-calculations/ /3
A working group at SPSP wrote a great paper about power analyses and common misperceptions. https://osf.io/9bt5s/ /4
This is a paper that describes how to use a Monte Carlo study to calculate power (using a simulation approach), which can be super helpful for very complex proposed analyses (e.g., growth curve models) http://dx.doi.org/10.1207/S15328007SEM0904_8 /5
I am a big fan of a safeguard power analyses, which help account for bias that exists in using an effect size estimate from a single sample, which likely does not accurately reflect the true population effect size. https://journals.sagepub.com/doi/full/10.1177/1745691614528519 /6
A power-calibrated effect size approach also helps adjust sample size estimates to better reflect the population effect size. https://psycnet.apa.org/doiLanding?doi=10.1037/met0000036 /7
And here is yet another approach for adjusting to better reflect the population effect size using a BUCCS shinyapp. https://journals.sagepub.com/doi/abs/10.1177/0956797617723724 /8
I am also a big fan of sequential analyses, since they help plan a study a priori to be highly powered, but then stop the study at a pre-determined interim point if certain criteria are met, thus reducing participant burden, saving time and money, etc. https://onlinelibrary.wiley.com/doi/abs/10.1002/ejsp.2023 /9
What are your favorite resources? /10
