In characterizing the statistical consequences of having a small number of subjects, the original article states, ‘using a two-sided sign test, demonstration of a consistent sign of an effect in six out of eight subjects corresponds to a p-value of 0.03125.’ Because of a coding error, this value is incorrect, and the correct p-value is 0.289 (two-sided) or 0.145 (one-sided). For completeness, we note that in the case of seven out of eight subjects, the corresponding p-value is 0.070 (two-sided) or 0.035 (one-sided), and in the case of eight out of eight subjects, the corresponding p-value is 0.008 (two-sided) and 0.004 (one-sided). More broadly, we note that this is just one simple approach towards assessing the reliability of findings across subjects, and more sophisticated approaches examining within-subject error and/or using quantitative values (as opposed to the mere sign) will likely have increased statistical power. The authors would like to acknowledge M.E. Nilsson from Stockholm University for catching the error, and apologize for any inconvenience caused.
Bibliographical notePublisher Copyright: