Actually, I think noise is a red herring, and that worrying about noise (instead of worrying about unmeasured confounders and incorrect causal structures) is how epidemiology got into this mess in the first place.