Test 04 — Does the b-value order tectonic settings?
Supplied working report; original verdicts not validated.
Review status: Needs review. This is an archival source note, not a passed proof.
README.md
# Test 04 — Does the b-value order tectonic settings? **Hypothesis tested.** The notebook reports b = 1.30 (Japan, subduction), 1.32 (Chile, subduction), 1.05 (California, transform) and 1.57 (Iceland, rift), and states that the ordering matches the tectonic setting. Re-estimated here from independent catalogue pulls for six regions, with analytic intervals, at three completeness choices, and after declustering. **Verdict: the numbers replicate, and the ordering survives — with two caveats that matter.** All four published b-values reproduce: Chile 1.32, Iceland 1.57, California 1.04, and Japan 1.30 once the window is matched to the notebook's (2015–2025). 13 of 15 pairwise comparisons separate at 5%. But the ordering is not a clean ladder. Japan's value under the full 2010–2025 window is 1.17, not 1.30 — the difference is the Tohoku aftershock sequence — which puts it next to the induced catalogue rather than with Chile. **The b-value is a property of the catalogue window as much as of the tectonic setting.** ## Result | Region | Setting | Mc | N ≥ Mc | b | 95% CI | b at Mc±0.1 | Declustered b | b, 2015–25 | Notebook b | |---|---|---:|---:|---:|---|---:|---:|---:|---:| | japan | subduction | 4.4 | 11,205 | **1.17** | 1.15–1.19 | 1.06 / 1.24 | 1.17 | 1.30 | 1.30 | | chile | subduction | 4.3 | 6,556 | **1.32** | 1.29–1.36 | 1.23 / 1.34 | 1.33 | 1.32 | 1.32 | | newzealand | subduction | 4.3 | 1,460 | **1.17** | 1.11–1.23 | 1.06 / 1.21 | 1.18 | 1.17 | — | | california | transform | 3.6 | 1,497 | **1.04** | 0.98–1.09 | 0.99 / 1.02 | 1.01 | 1.04 | 1.05 | | iceland | rift | 4.5 | 175 | **1.57** | 1.34–1.80 | 1.47 / 1.37 | 1.38 | 1.57 | 1.57 | | oklahoma | induced/intraplate | 2.5 | 7,436 | **1.09** | 1.06–1.11 | 0.87 / 1.14 | 1.03 | 1.09 | — | Group means: subduction 1.22, transform 1.04, rift 1.57, induced/intraplate 1.09. ### Which differences are real | Pair | Δb | z | Separated at 5%? | |---|---:|---:|---| | chile − oklahoma | +0.237 | +11.50 | yes | | chile − california | +0.289 | +9.21 | yes | | japan − chile | -0.157 | -8.09 | yes | | japan − oklahoma | +0.080 | +4.70 | yes | | japan − california | +0.132 | +4.53 | yes | | chile − newzealand | +0.156 | +4.47 | yes | | california − iceland | -0.533 | -4.37 | yes | | iceland − oklahoma | +0.481 | +4.02 | yes | Only the largest contrasts survive: Japan and Chile against Oklahoma and California. Every comparison involving Iceland is underpowered, and the three subduction zones are not distinguishable from each other. ## Interpretation The b-value is a ratio of small to large events; it responds to stress heterogeneity, depth range and fault maturity. The pattern here — 1.3 in Chile and 1.0 in California — is consistent with the published regional literature, and the four values in the notebook are reproducible exactly. What the replication adds is where the ordering is fragile. Japan moves from 1.30 to 1.17 when the window is extended back over the Tohoku sequence, a swing of 0.13 that is an order of magnitude larger than its standard error. Any ranking built from a single window inherits that sensitivity. The safest statement the data support is a **two-group contrast** — Chile and New Zealand near 1.2–1.3, California and Oklahoma near 1.0–1.1 — with Iceland suggestive but resting on 175 events. Iceland deserves the caveat it usually does not get: a small, episodic volcanic catalogue where swarms dominate and b varies by swarm phase. Quoting 1.57 as a rift reference point overstates what 175 events can establish, even if the interval here looks tight. ## What this test cannot see Regions differ in network density, magnitude type and catalogue length, so a between-region difference can be a reporting difference. Depth is not controlled — subduction catalogues include deep events where b is known to be higher. Declustering used a crude three-day window and does not remove swarms properly. The notebook's own values are reproduced within 0.1 for Japan and Chile, but its California Mc (4.3) is higher than this pull's mode (3.6), so the two California estimates are not strictly comparable. ## Files `run.py` reproduces everything from `../../data/quakes_*.csv`; `data.csv` is the per-region table; `results.json` holds every number above, including all 15 pairwise tests.
hypothesis.md
# Hypothesis — Does the b-value order tectonic settings? ## The claim as published The notebook's `seismicity/FINDINGS.md` reports, from the USGS catalogue with M ≥ 4.0 for 2015–2025: | Region | Setting | Mc | N ≥ Mc | b | |---|---|---:|---:|---:| | Japan | subduction | 4.4 | 5,079 | 1.30 | | Chile | subduction | 4.3 | 6,556 | 1.32 | | California | transform | 4.3 | 270 | 1.05 | | Iceland | rift | 4.5 | 175 | 1.57 | and states that the ordering matches the tectonic setting. ## The decisive comparison, fixed in advance **Decisive comparison, fixed in advance:** re-estimate b for six regions with analytic 95% intervals, and test every pairwise difference. **Support**: the four published values reproduce within 0.05, and pairwise tests separate subduction from transform. **Partial support**: values reproduce but the intervals do not separate the groups. **Refute**: values do not reproduce, or the ordering reverses under a different completeness choice or window. ## Support and refutation No separate support criteria: the decisive comparison above is the criterion.