Is principal components analysis necessary to characterise dietary behaviour in studies of diet and disease?