I’m not talking about blinding, I’m just talking about randomizing. That’s right, in areas with obvious confounders like class, baby health, and maternal stress level, and relatively small differences in outcomes between the groups anyway, I don’t think correlational data is worth much.
Having parented a difficult-to-feed baby and having tried everything I could think of to get calories into her, I’m quite sure that even parents who start out willing to follow a given recommendation quickly change their mind if things don’t seem to be going well. (If not, you’re selecting for parents who are willing to prioritize following instructions over their baby’s health, which certainly gets you a different population than is typical.)
I’m not talking about blinding, I’m just talking about randomizing. That’s right, in areas with obvious confounders like class, baby health, and maternal stress level, and relatively small differences in outcomes between the groups anyway, I don’t think correlational data is worth much.
Having parented a difficult-to-feed baby and having tried everything I could think of to get calories into her, I’m quite sure that even parents who start out willing to follow a given recommendation quickly change their mind if things don’t seem to be going well. (If not, you’re selecting for parents who are willing to prioritize following instructions over their baby’s health, which certainly gets you a different population than is typical.)