Interface StatTestResult

Bootstrap paired significance test for comparing two experiment runs. Pure TypeScript, no external dependencies.

interface StatTestResult {
    baselineMrr: number;
    candidateMrr: number;
    confidenceInterval95: [number, number];
    mrrDelta: number;
    pValue: number;
    recommendation: "accept" | "reject" | "inconclusive";
}

Properties

baselineMrr: number

candidateMrr: number

confidenceInterval95: [number, number]

95% confidence interval on the MRR delta [lower, upper].

mrrDelta: number

pValue: number

One-tailed p-value: P(candidate ≤ baseline | H0: no difference).

recommendation: "accept" | "reject" | "inconclusive"

"accept" (p < 0.05 and delta > 0), "reject" (p < 0.05 and delta ≤ 0), or "inconclusive".