Neuroscientists Too Often Exceed Chance Levels Only By Chance


Many neuroscientific studies are based on relatively small sample sizes, but with large data sets — together, that makes the studies rife for generating apparently meaningful results that are actually just being generated by chance background “noise,” according to a new study discussed by Neuroskeptic in Discover. And many neuroscientific studies have not been correcting for these possible biases.

Machine Learning: Exceeding Chance Level By Chance (Discover, January 18, 2015)


  1. Which is because a lot of biomedical scientists don’t know sh*t about statistics and don’t see a problem in increasing the numbers until but not over the point when they reach the magical p<0.05. And if you actually design and perform the experiments the way you should your supervisor complains why you get so few positive results. Welcome to reality.

    Report comment