Skip to content

Sequential evidence ​

This page defines the mean of e-values, the rule that combines them under any dependence. It follows the conventions of statistics capabilities: the RangeError rules, the accuracy of 1e-11.

Status: partly built. eValueAverage exists in @agent-nexus/driftbench-stats; the other functions are planned.

E-values ​

An e-value is a non-negative number whose expected value is at most 1 when the null is true. A large e-value is evidence against the null. The function below takes e-values as plain numbers: each one is at least 0 and finite. It is defined in packages/stats/src/evalues.ts.

eValueAverage ​

ts
eValueAverage(evalues: number[]): number
  • Definition. The arithmetic mean: the sum of the values divided by their count. The mean of e-values is an e-value under any dependence between them.
  • Returns. A number, at least 0.
  • Throws. A RangeError for an empty array and for a value that is negative, NaN or infinite.
  • Example. eValueAverage([2, 4]) is 3. eValueAverage([0, 0, 3]) is 1. eValueAverage([0.5, 1.5, 2, 10]) is 3.5. eValueAverage([0.1, 0.2, 0.3, 40]) is 10.15.

MIT. driftbench is made by Nexus.