W2: When does more testing stop resolving the engineering question?

[Zawwarsami16]

A successful demonstration does not, by itself, establish dependable service. My W2 paper examines that distinction through documented AAG and EMALS development milestones and reproducible analytical examples.

One example makes the test-design question concrete: changing elapsed exposure separates cycle and time effects in an additive model, yet can leave time and interaction effects indistinguishable in a larger model family.

What is the smallest additional observation that would justify the stronger conclusion? When should an evaluator seek another measurement, recover missing records, change the test conditions, or collect more repetitions?

Counterexamples and earlier work are welcome. The numerical examples are synthetic; they are not measurements of carrier performance.

• W2 reading folder and PDF
• Zenodo · PhilPapers
• Official paper page
• Detailed evidence and corrections — issue #19