Read on Twitter
Aaaaargh. My half yearly scream of just *please* show me (grant reviewer) you (applicant) have done a power analysis. I honestly don& #39;t want much - just something to show you have thought through whether your data gathering + analysis experiment has some hope in working.
It& #39;s just not good enough to either (a) not talk about it at all (they will send it to reviewers like me) or (b) imply that you get a pass because you will be doing nearly all the cases in country / hospital / region or the maximum mice you can fund.
For (b) I firstly don& #39;t buy you will be able to do all (no one ever is) and secondly the point of a power calculation is whether you have some hope of making an inference. If you can& #39;t get enough samples to have that hope then ... its a pretty pointless study.
As I& #39;ve often said, this is about showing that you (the applicant) aren& #39;t fooling *yourself* about what is inferable or not. You can always & #39;fudge& #39; a power calculation (via setting an effect size) it& #39;s just the honesty about the effect size with yourself I want to see
I would literally give more marks for power calculations done on the back of the envelope and using your mobile phone to take a picture of the envelope. It& #39;s not that hard to do a power calculation. it is a sentence.
"Assuming the effect size of the treatment is similar to Wibble& #39;s study of Wobble (ref), we calculate that there is a 80% chance of us observing an effect with this sample size". Not "we will do 50 patient samples which is the maximum yearly number in our hospital".
