Must read: Emanuel Wyler recommends a paper on a statistical fallacy
Measurements with small sample sizes can become ‘statistically significant’ due to the combination of small sample numbers and measurement error. This produces effects that are not present in the system under scrutiny.