June 4, 2019, Vol 321, No. 21, Pages 2045-2142
The Importance of Predefined Rules and Prespecified Statistical AnalysesDo Not Abandon Significance
John P. A. Ioannidis, MD, DSc
JAMA. 2019;321(21):2067-2068. doi:10.1001/jama.2019.4582
In this Viewpoint, John Ioannidis argues against abandoning the notion and language of statistical significance, which has been proposed as a means to diminish oversimplistic interpretations of clinical research. A significance filter in some form is essential for distinguishing signal from noise, he writes, and emphasizes that predefined study design choices, prespecified statistical analyses, transparent and documented deviations from either, and improvement in researchers’ statistical numeracy can minimize overly subjective interpretations of whatever significance measure is used.