The negative class: what a passing case proves

Label lab: L3.3 released 47 cases and every one is a failure, so a grader that answers fail to everything has read 100.0% on the dataset and 11 of 11 on the release gate for six weeks. Put L2.1's 16 recorded passes back and the same constant falls to 74.6% on 63 cases and 40.7% on the gate, with no check changed. The passes fill 3 of 3 slices and 0 of 9 mechanisms, so the only key they supply is the request envelope — days 3–30, amounts 15.00–52.20 — which excludes 23 of 47 failures and, fitted on half the passes, flags 4 of the other 8. Zero false alarms in 16 passes bounds the rate at 18.8%; 5% needs 60, which is 44 more than exist.

This Evals Engineering lesson is part of a paid plan. Sign in with a plan that includes this course to read it.