Thread by @PompanoLab

The Pompano lab group meeting topic today: #poweranalysis. PhD student Alex opened with prompt: "How do you choose how many samples to run when testing the effect of X on Y?"

Initial answers:
* Always n = 5
* Whatever # is convenient
* Lots of blank faces

Resources below.

For some reason we don't teach power analysis in Chemistry or most Biology programs (undergrad), nor in grad Chem or most grad Bio. Whereas in social sciences, this is intro-level stuff. #needtoupdatethecurriculum

Yet, we need to learn it to ensure that we can accurately detect small changes between groups in our experiments. In fact, @NIH mandates power analysis to justify proposed animal or human experiments.

@NIH I am no statistician, but fortunately there are excellent resources online for the average biomedical researcher to get up to a functional level for conducting power analysis in < 1 day of effort. Once you know how, it takes 5 minutes.

@NIH Here are the intro materials my group found useful:

1. Handbook of Biological Statistics, by John McDonald
biostathandbook.com/power.html

@NIH 2. A Researcher’s Guide to Power Analysis, by Anne-Therese Hunt
research.usu.edu//irb/wp-conten…

@NIH 2b. Dr. Hunt and @USUAccess even provided powerpoint slides that you can use as a starting point to teach this to your group. We started with these today, then added our own examples.

research.usu.edu/irb/analysis/

@NIH @USUAccess 3. Of course, real experiments do not always follow the introductory textbook examples. I really appreciate this blog entry from Simon Bate:

"How to decide your #samplesize when the power calculation is not straightforward"
nc3rs.org.uk/news/how-decid…

@NIH @USUAccess And finally, a free online calculator. Once you understand the basics from the resources above, using this calculator is a breeze. You can plan your next experiment in minutes. We practiced using this as a group, and now my students are empowered.
powerandsamplesize.com

/end