We greatly appreciate Cristian Antony Ramos-Vera's interest in our article; in his letter he highlights the virtues of using Bayes factor (BF) as an alternative to the traditional dichotomous interpretation of hypothesis testing, and his analysis provides more robust support for our findings.1

Frequentist statistics almost entirely dominate medical research. The average reader has interiorized the concepts of hypothesis testing, P value, and statistical significance. The limitations of frequentist statistics and the problems with their interpretation have been widely discussed2 and, in addition, repeated appeals have been made to include Bayesian statistics in biomedical research.3 While it is true that Bayesian statistics allow a more natural and intuitive interpretation, the reality is that their use is not widespread and most readers do not understand them.

Hoekstra et al.4 performed a reanalysis of 36 articles with negative results and calculated the BF. The smallest BF was 2.42 (observed data are 2.42 times more probable under the null hypothesis) and the largest, 560.9. A key point is that there was a poor correlation between the P value and the BF. A high P value may have been present in studies with little evidence in favor of the null hypothesis (low BF) or in studies with strong evidence (high BF). This allows us to assert that the BF intuitively communicates the probative strength of the hypothesis; therefore, we, like Dr Ramos-Vera, recommend that this should routinely be included in scientific articles.