Read on Twitter
Here's my latest attempt at demonstrating the despondency of responder counting in exercise science. First, I hope it's clear by now that looking at a plot like this in a single group tells us nothing about individual response differences.
So sometimes we see comparisons of responder or non-responder counts between 2 or more groups. For example, there appear to be 10 non-responders (change < -2 units) in this group of 30 people.
But in this group of 30 people there appears to be only 1 non-responder (change < -2 units). So do we have a difference between these groups in terms of individual "responsiveness" or individual "trainability"?
Absolutely not. What if I told you that those two groups were actually paired data from the same people. Every person improved by 5 units. There are no individual diffs in response at all. The difference in non-responder counts is simply due to difference in mean change.
So, again, all these responder count comparisons could be are low powered proxy analyses of group differences in mean response. This is nothing to do with precision exercise medicine.
