Scanning the paper the effect seems surprisingly weak (and has p=0.045!) and weirdly dependent on family size (it looks as thought he difference only exists in

Scanning the paper the effect seems surprisingly weak (and has p=0.045!) and weirdly dependent on family size (it looks as thought he difference only exists in larger families). While the results seem believable I'm not convinced by the results as presented.