Methodology

How the disclosure-day null was measured — and bounded.

Six benchmark models, two event-date anchors, three primary windows, an exhaustive placebo battery, an AI-loading cross-section, a GARCH refit, and a pre-specified power analysis. The design is built to find a Dell-specific effect if one existed; the admissible claim is strictly “no detectable effect, MDE .”

Open any method in Plain English or Academic detail

The estimators.

Each button opens a three-tab modal: a Plain-English explanation, a worked example bound to this study’s numbers, and the academic formula with citations. No number is hardcoded — the worked examples resolve against the canonical JSON.

Design parameters

The estimation window, anchors, and windows.

Estimation window

trading days; -day gap

Estimation σ

daily abnormal-return volatility

Primary anchor

T0 = 

co-equal sensitivity anchor T0 = 

Primary windows

wider windows confound-flagged

Event-date discipline. EDGAR acceptances were post-close (PRE 14A , DEFA14A ET on ), so May 4 and May 5 are reported co-equal until the Business Wire wire-minute is pinned; no anchor is privileged to flatter the result.

Multiple testing & power

The minimum detectable effect is always reported.

Multiple-testing control

  • Benjamini-Hochberg is applied within a pre-specified -cell primary family, kept separate from the -cell exploratory grid, so the headline pBH =  is not inflated by the wider grid.
  • Wide windows that overlap identified confounders are confound-flagged and excluded from headline inference.

Power limits (honest)

  • MDE at 80% power (α = 0.05, two-sided): at , at , at .
  • The TOST at ±2 pp returns p ≈  — it does not affirm equivalence. This is a power-limited null, not an equivalence-confirmed one.
  • With only donors the smallest attainable permutation p is ; placebo-in-space characterizes Dell’s position, it cannot reject at 5% until the pool reaches ≥ .

The mechanism

The cross-section is the diagnostic.

The AI-loading cross-section regresses each peer’s May-5 CAR on its AI-beta. If the apparent May-5 significance were a Dell-specific signal, Dell would sit far off the line. Instead it sits on it. The full panel (N = ) gives slope , R² , p = ; screening Arista (an earnings confound, screen #2) gives N = , slope , R² , p =  — the gradient strengthens. Preliminary at N = 7/8, pending the ≥19-donor export. The interactive scatter and per-firm methodology are on the Stress tests page.