Registered protocol 36d019b175c8 (Lab clustered decile inference v1) ·
method 4cfe94a4f4ec7a7f · registry-first · built 2026-07-27 01:35 UTC
1 · G<8 UNDERPOWERED threshold retained. The wild-cluster
(Rademacher) small-G remedy is computed beside every cluster CI; 8 matches the smallest live lens;
changing thresholds after results exist is forbidden by our own discipline.
2 · Percentile CIs retained for v1. Bootstrap-t is deferred —
it needs studentization machinery not yet exercised in this codebase; revisit at v5.3 under its own registered spec.
| Horizon | Mean spread | Ticker-cluster CI (primary) | Wild-cluster CI (small-G remedy) | Naive CI (comparison) | dates/tickers/obs | Badge |
|---|---|---|---|---|---|---|
| k=1 | +0.05% | (-0.42%, +0.54%) | (-0.46%, +0.56%) | (-0.65%, +0.70%) | 47d / 65t / 800o | DESCRIPTIVE |
| top decile mean (secondary) | -0.10% | (-0.39%, +0.19%) | ||||
| bottom decile mean (secondary) | -0.15% | (-0.56%, +0.26%) | ||||
| k=5 — PRIMARY | +0.18% | (-1.45%, +1.81%) | (-1.59%, +1.94%) | (-1.29%, +1.68%) | 43d / 63t / 736o | DESCRIPTIVE |
| top decile mean (secondary) | -1.22% | (-2.22%, -0.23%) | ||||
| bottom decile mean (secondary) | -1.40% | (-3.09%, +0.30%) | ||||
| k=20 | -1.99% | (-6.28%, +2.55%) | (-6.63%, +2.77%) | (-5.40%, +1.16%) | 27d / 60t / 480o | DESCRIPTIVE |
| top decile mean (secondary) | -5.43% | (-9.17%, -1.51%) | ||||
| bottom decile mean (secondary) | -3.44% | (-7.09%, +0.07%) | ||||
Estimator note: the published Lab panel reports per-rebalance hold-to-next spreads; this preview uses per-signal-date k-day forward-return spreads because ticker identity is required for clustering — the two are stated side by side, never blended. Same tickers appear in deciles on many dates; the ticker-clustered CI absorbs that dependence, the naive date-resample CI does not. Percentile intervals; bootstrap-t deferred (decision 2 above).
| Context cell | Spread estimate | Ticker-cluster CI | n/G | Badge |
|---|---|---|---|---|
| k=1 | +0.39% | (-0.04, +0.86) | 848/67 | DESCRIPTIVE |
| k=5 (base) | +0.77% | (-0.87, +2.39) | 784/64 | DESCRIPTIVE |
| trend>=0 (k=5) | empty — SPY history begins 2026-02-02; a trailing-120-trading-day window is not computable at these signal dates (own-data rule) | |||
| trend<0 (k=5) | empty — SPY history begins 2026-02-02; a trailing-120-trading-day window is not computable at these signal dates (own-data rule) | |||
| vol-high (k=5) | +0.33% | (-2.59, +3.55) | 416/62 | DESCRIPTIVE |
| vol-low (k=5) | +1.29% | (-0.28, +3.12) | 368/55 | DESCRIPTIVE |
| k=20 | -1.29% | (-5.87, +3.21) | 528/61 | DESCRIPTIVE |
Sign held in 4/5 computed cells · CI excluded zero in 0/5 · breaks in: k=20 · UNDERPOWERED cells: none. Every cell ledger-counted. 449 secondary cells ledgered registry-wide; expected false positives at alpha=0.05: 22.45. Investment implication: none established — no buy, sell, or alpha conclusion is supported by this page.
Protocol 36d019b175c8 locked 2026-07-27 ·
primary endpoint: ticker-clustered bootstrap 95% CI on the top-minus-bottom decile spread of return_5d (k=5), forward-OOS window only
Run recorded: registry line 5bcdfb6720e0bbcf · chain verified at build time.
Registered before computation on real data.