Response to ‘Why P values are not measures of evidence’ by D. Lakens

In our recent opinion article in TREE [1], we suggested that P-values should be interpreted gradually, rather than dichotomously, and with reference to degrees of evidence instead of to statistical significance. Gradual interpretation is consistent with advice given in several statistical text books [2,3] and a recent publication [4], but our suggestion of using language of evidence in relation to P-values has sparked vigorous discussion. This is not surprising, given that ‘evidence’ is not a mathematically defined concept.