This makes sense, to be, so, perhaps, I'm confused by how he would then go on to talk about the individual values that make up the pair. Is that, then, exactly

This makes sense, to be, so, perhaps, I'm confused by how he would then go on to talk about the individual values that make up the pair. Is that, then, exactly his point? That a random sample of a group can't be modeled like a group of random samples, yet this is what's being done?