Sure, but when the nominal increase is 18% and the selected group only shows 3% there's a big question about group selection. If you can cherry pick groups and

Sure, but when the nominal increase is 18% and the selected group only shows 3% there's a big question about group selection. If you can cherry pick groups and variables then you can use phenomena like Simpson's paradox to prove pretty much anything with statistics.