Tuesday, March 18, 2008

A matter of significance

The use of the word "significant" in scientific writing poses a unique problem because it is implicitly ambiguous in the term "statistically significant".  While the dictionary definition of "significant" does not necessarily invoke statistical measure, and its synonyms include "notable" and "important", its use in a nonstatistical sense in a scholarly article amounts to a misuse.  

This is especially true when the comparison being made could indeed be tested for statistical significance. In such a case, the "statistical" modifier should be understood, because a statistical test should be undertaken during any serious analysis.

During broad introductions, the use of "significant" in a nonstatistical sense may be appropriate, although its use should be discouraged. Elsewhere in scientific writing, it should absolutely be avoided unless a statistical demonstration of significance is offered. This guideline will disambiguate a term for which there are an ample number of synonyms to be used in cases where statistical significance cannot be shown.

contributed by Dr. Ron McTaggart-Cowan
Numerical Weather Prediction Section
Meteorological Service of Canada