The thing is, small sample sizes, be it as they may, are still data, and drawing inferences off of them is the only thing we can do given the nature of our inability to experiment.