LOCAL ANALYSIS · COHORT DATA STAYS ON-DEVICE

WANTED
Score Lab.

Enter one record per independent environment. The calculator reproduces the benchmark’s Kaplan–Meier restricted mean survival score, refuses unsupported 10,000-hour extrapolation, and reveals censoring sensitivity, tail support, and leave-one-environment-out influence.

← PROTOCOL KIT
WANTED SCORE85.2NORMALIZED RMST / 100
BOOTSTRAP 95% CI62.0–100.010,000 DETERMINISTIC ENVIRONMENT RESAMPLES
RETENTION AT 10K60.0%Ŝ(10,000)
STATISTICAL STATUSPROVISIONAL5/20 SITES · 42,600/10,000 H
COHORT INPUTONE ROW PER INDEPENDENT ENVIRONMENT
ENVIRONMENTRESIDENT HOURSOUTCOME
5 ENVIRONMENTS · 2 EVENTS · 0 TERMINAL REVIEWS · 42,600 HOURS
ROBUSTNESS DISCLOSUREHow fragile is W?
METHOD CONTRACT ↗
CENSORING ENVELOPE85.2–85.2W WIDTH 0.0 · 0 EARLY EXITS
TAIL SUPPORT30 REJECTED AT 10K · 3 RETAINED
MAX LEAVE-ONE-OUT SHIFT±10.80ENV-004
UNSUPPORTED EXCLUSIONS0OF 5 LEAVE-ONE-OUT RUNS
PESSIMISTICW 85.20

Every early non-rejection exit fails at its last observed hour.

≤
PRIMARYW 85.20

Preregistered Kaplan–Meier treatment of independently censored exits.

≤
OPTIMISTICW 85.20

Every non-rejection exit is retained through the 10K horizon.

ENVIRONMENT REMOVEDW WITHOUT ENVIRONMENTSHIFT
ENV-00496.000+10.800
ENV-00181.500-3.700
ENV-00381.500-3.700
ENV-00581.500-3.700
ENV-00285.500+0.300

Bounds are deliberate stress scenarios, not replacement estimators. Leave-one-out influence is descriptive and must be reviewed when it exceeds the preregistered threshold. Reproduce an official claim →

STATISTICAL ELIGIBILITY IS NOT CERTIFICATION.A qualifying cohort still needs complete intervention disclosure, robustness review, passed safety gates, preregistration, and independent audit.