Biotech Values

[JohnWayne]

HALO/OT: power and alpha.

I’m no stats expert, but I have a basic understanding and love learning more. Here’s my guess as to why the increase in alpha from 0.04 to 0.05 didn’t result in a larger increase in power.

Take this graphical schematic showing the relation between alpha and beta/power as a starting point. https://www.animatedsoftware.com/statglos/sgrelabp.htm

This shows a hypothetical study designed with a power of 0.26. Increasing the power of the study would effectively be sliding the bottom distribution even further to the right, such that the dark green area under the curve to the right of the critical value was e.g. 0.92 for 92% power. Then, imagine changing the alpha on the top graph from 0.04 to 0.05 - the critical value would move very slightly to the left, which would increase the area under the curve that represents power. But the power would only increase a tiny amount, as it’s all coming in the tail of the curve which represents very little area