Section 4 of 6
On the Appropriate Use of Prediction Intervals
Ibnu Noufal Kambitta Valappil, Karuppasamy Govindasamy, Gavoutamane Vasanthi, Masilamani Elayaraja, Cain C. T. Clark, Koulla Parpa, Borko Katanic, Hüseyin Şahin Uysal, Hassane Zouhal, and Urs Granacher · about 1 minutes
We share the colleagues’ enthusiasm for the reporting of 95% prediction intervals (PIs) alongside confidence intervals [23]. We note, however, that the PIs reported in the commentary’s sensitivity analyses (for example, − 1.59 to 2.99 for dynamic balance, − 1.44 to 0.50 for COD, − 0.91 to 1.53 for Sargent jump) reflect expected variability across contexts rather than absence of effect. They are therefore not grounds for dismissing the substantive conclusions.