Statistical Significance

A result is called statistically significant when it would be unlikely to arise by chance alone if there were truly no effect, judged by a preset cutoff, most often a p-value below 0.05 or a 95 percent confidence interval that does not include no effect. It is a statement about chance, not about how large or important an effect is.

Origin

The convention of a 5 percent significance level traces largely to the statistician Ronald Fisher's work in the 1920s. It became the standard test of research results in medicine and the social sciences through the 20th century.

Significance in Asbestos Context

Statistical significance is argued over constantly in asbestos litigation. Defense experts often say that a study showing a higher cancer rate among exposed workers is not significant and therefore does not show an effect. Plaintiffs' experts respond that a non-significant result is not proof of no effect, especially in small studies of rare diseases like mesothelioma, where a study may be too small to detect real risk, and that consistent results across many studies matter more than any single test. Scientists increasingly warn against treating 0.05 as a bright line. Courts differ in how much weight they give it; in a 2011 securities case, Matrixx Initiatives v. Siracusano, the U.S. Supreme Court observed that medical professionals do not always rely on statistical significance to infer causation. Significance also says nothing about whether an effect is large: a tiny effect can be significant in a huge study, and a large one can miss significance in a small study.

Example Sentence

The defense argued the study's 40 percent increase in laryngeal cancer was not statistically significant, while the plaintiff's expert noted that the study had few cases.

Related Terms

P-Value, confidence interval, Relative Risk, Epidemiology, Bradford Hill Criteria, Daubert Standard, General Causation

Notes

• "Not statistically significant" does not mean "proven safe"; it may mean the study was too small.
• Ask how many cases a study had; rare diseases often produce small numbers.