Thinking, Fast and Slow

by Daniel Kahneman

Using a sufficiently large sample is the only way to reduce the risk. Researchers who pick too small a sample leave themselves at the mercy of sampling luck.