Power of negative thinking M.S. Avidan, T.S. Wildes British Journal of Anaesthesia Volume 114, Issue 1, Pages 3-5 (January 2015) DOI: 10.1093/bja/aeu263 Copyright © 2015 The Author(s) Terms and Conditions
Fig 1 A statistical argument suggesting that much of published positive research is false. This figure illustrates how a high proportion of positive findings could be false-positive findings. The conclusion is based on the assumption that only a low number of hypotheses (e.g. 10%) are true. Of the 100 true hypotheses, about 80 would be detected as positive based on 80% power of the study design. In addition, ∼45 of the 900 false hypotheses would be incorrectly detected as positive based on a P-value threshold of 5%. If these assumptions were correct, it would also follow that very few negative findings would be likely to be false-negative results. With reference to the figure, only 2.3% [20/(855+20)] of the negative findings would be false negatives. Modified from17, with permission from Alex Tabarrok. British Journal of Anaesthesia 2015 114, 3-5DOI: (10.1093/bja/aeu263) Copyright © 2015 The Author(s) Terms and Conditions