Model benchmarking with a statistical grading scale

This paper presents a method for deriving a grading scale for model benchmarking and applies the technique to the benchmarking of the SABER methodology for lethality assessments. The method proposes a grading scale based on the level of acceptance of a goodness-of-fit test. The probability α of rejecting a hypothesis when it is true, as derived from an observed value of a test statistic, is often referred to as the level of acceptance of the test. The hypothesis to be tested here is that the SABER model is valid for a specific application, with α being used as a basis to define various grading levels as an indicator of benchmark success