There seems to be a glaring problem in statistics in general that large scale statistics may completely hide small group effects. I.e., if there is something th

There seems to be a glaring problem in statistics in general that large scale statistics may completely hide small group effects. I.e., if there is something that really benefits from X for 5%, but doesn't benefit from X for 95%, statistics will make it look like it doesn't benefit anybody. Yet that 5% group may absolutely be relevant. This always sits in the back of my head when I read things like: "Children are not influenced by their parents but by their peers", or "twin study proves this". But I haven't seen this discussed much or brought up, even though it sounds a lot like what you're saying. Is this relevant? Is there a name for this? Is there some reason as to why I shouldn't worry about this that I'm missing?