Comment by rrtwo

What is the generalised rule/case where small sample sizes are sufficient?