2026-04-24 05:51:30 +00:00
\relax
\providecommand \hyper @newdestlabel[2]{ }
\providecommand \HyperFirstAtBeginDocument { \AtBeginDocument }
\HyperFirstAtBeginDocument { \ifx \hyper @anchor\@ undefined
\global \let \oldnewlabel \newlabel
\gdef \newlabel #1#2{ \newlabelxx { #1} #2}
\gdef \newlabelxx #1#2#3#4#5#6{ \oldnewlabel { #1} { { #2} { #3} } }
\AtEndDocument { \ifx \hyper @anchor\@ undefined
\let \newlabel \oldnewlabel
\fi }
\fi }
\global \let \hyper @last\relax
\gdef \HyperFirstAtBeginDocument #1{ #1}
\providecommand \HyField @AuxAddToFields[1]{ }
\providecommand \HyField @AuxAddToCoFields[2]{ }
\citation { Homola2023}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { Kanamori1977}
2026-04-24 05:51:30 +00:00
\citation { Stoupel1990,Urata2018}
\citation { Homola2023}
\citation { Bretherton1999}
\citation { Potgieter2013}
\citation { Odintsov2006,Tavares2011}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 1} Introduction} { 5} { section.1} \protected @file@percent }
\newlabel { sec:intro} { { 1} { 5} { Introduction} { section.1} { } }
2026-04-24 05:51:30 +00:00
\citation { USGS2024}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { Kanamori1977}
2026-04-24 05:51:30 +00:00
\citation { SIDC2024}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 2} Data} { 6} { section.2} \protected @file@percent }
\newlabel { sec:data} { { 2} { 6} { Data} { section.2} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 2.1} Cosmic-Ray Flux: NMDB Neutron Monitors} { 6} { subsection.2.1} \protected @file@percent }
\newlabel { sec:nmdb} { { 2.1} { 6} { Cosmic-Ray Flux: NMDB Neutron Monitors} { subsection.2.1} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 2.2} Seismic Activity: USGS Earthquake Catalogue} { 6} { subsection.2.2} \protected @file@percent }
\newlabel { sec:usgs} { { 2.2} { 6} { Seismic Activity: USGS Earthquake Catalogue} { subsection.2.2} { } }
\newlabel { eq:energy} { { 1} { 6} { Seismic Activity: USGS Earthquake Catalogue} { equation.2.1} { } }
\citation { Bartlett1946}
2026-04-24 05:51:30 +00:00
\citation { Bretherton1999}
\citation { Theiler1992,Schreiber2000}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 2.3} Solar Activity: SIDC Sunspot Number} { 7} { subsection.2.3} \protected @file@percent }
\newlabel { sec:sidc} { { 2.3} { 7} { Solar Activity: SIDC Sunspot Number} { subsection.2.3} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 3} Methods} { 7} { section.3} \protected @file@percent }
\newlabel { sec:methods} { { 3} { 7} { Methods} { section.3} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.1} Cross-Correlation at Lag $ \tau $ } { 7} { subsection.3.1} \protected @file@percent }
\newlabel { sec:xcorr} { { 3.1} { 7} { Cross-Correlation at Lag $ \tau $ } { subsection.3.1} { } }
\newlabel { eq:xcorr} { { 2} { 7} { Cross-Correlation at Lag $ \tau $ } { equation.3.2} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.2} Effective Degrees of Freedom} { 7} { subsection.3.2} \protected @file@percent }
\newlabel { sec:neff} { { 3.2} { 7} { Effective Degrees of Freedom} { subsection.3.2} { } }
\newlabel { eq:neff_ bartlett} { { 3} { 7} { Effective Degrees of Freedom} { equation.3.3} { } }
\newlabel { eq:neff_ breth} { { 4} { 7} { Effective Degrees of Freedom} { equation.3.4} { } }
2026-04-24 05:51:30 +00:00
\citation { Schreiber2000}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.3} Surrogate Significance Tests} { 8} { subsection.3.3} \protected @file@percent }
\newlabel { sec:surrogates} { { 3.3} { 8} { Surrogate Significance Tests} { subsection.3.3} { } }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 3.3.1} Phase Randomisation} { 8} { subsubsection.3.3.1} \protected @file@percent }
\newlabel { sec:phase} { { 3.3.1} { 8} { Phase Randomisation} { subsubsection.3.3.1} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\newlabel { eq:phase} { { 5} { 8} { Phase Randomisation} { equation.3.5} { } }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 3.3.2} IAAFT Surrogates} { 8} { subsubsection.3.3.2} \protected @file@percent }
\newlabel { sec:iaaft} { { 3.3.2} { 8} { IAAFT Surrogates} { subsubsection.3.3.2} { } }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 3.3.3} Block-Bootstrap Surrogates} { 8} { subsubsection.3.3.3} \protected @file@percent }
\newlabel { sec:blockbootstrap} { { 3.3.3} { 8} { Block-Bootstrap Surrogates} { subsubsection.3.3.3} { } }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 3.3.4} Global $ p $ -Value} { 8} { subsubsection.3.3.4} \protected @file@percent }
\newlabel { sec:pvalue} { { 3.3.4} { 8} { Global $ p $ -Value} { subsubsection.3.3.4} { } }
\newlabel { eq:pglobal} { { 6} { 8} { Global $ p $ -Value} { equation.3.6} { } }
2026-04-24 05:51:30 +00:00
\citation { HP1997}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { RahnUhlig2002}
2026-04-24 05:51:30 +00:00
\citation { Cleveland1990}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 3.3.5} GPU Acceleration} { 9} { subsubsection.3.3.5} \protected @file@percent }
\newlabel { sec:gpu} { { 3.3.5} { 9} { GPU Acceleration} { subsubsection.3.3.5} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\newlabel { eq:matmul} { { 7} { 9} { GPU Acceleration} { equation.3.7} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.4} Solar-Cycle Detrending} { 9} { subsection.3.4} \protected @file@percent }
\newlabel { sec:detrend} { { 3.4} { 9} { Solar-Cycle Detrending} { subsection.3.4} { } }
\newlabel { eq:hp_ lambda} { { 8} { 9} { Solar-Cycle Detrending} { equation.3.8} { } }
\citation { Kraskov2004}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.5} Partial Correlation Analysis} { 10} { subsection.3.5} \protected @file@percent }
\newlabel { sec:methods:partialcorr} { { 3.5} { 10} { Partial Correlation Analysis} { subsection.3.5} { } }
\newlabel { eq:partialcorr} { { 9} { 10} { Partial Correlation Analysis} { equation.3.9} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.6} Nonlinear Dependence Tests} { 10} { subsection.3.6} \protected @file@percent }
\newlabel { sec:methods:nonlinear} { { 3.6} { 10} { Nonlinear Dependence Tests} { subsection.3.6} { } }
\@ writefile{ toc} { \contentsline { paragraph} { Spectral coherence.} { 10} { section*.2} \protected @file@percent }
\newlabel { eq:coherence} { { 10} { 10} { Spectral coherence} { equation.3.10} { } }
\@ writefile{ toc} { \contentsline { paragraph} { Mutual information.} { 10} { section*.3} \protected @file@percent }
\newlabel { eq:mi} { { 11} { 10} { Mutual information} { equation.3.11} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.7} Geographic Localisation Scan} { 10} { subsection.3.7} \protected @file@percent }
\newlabel { sec:geo} { { 3.7} { 10} { Geographic Localisation Scan} { subsection.3.7} { } }
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\citation { Benjamini1995}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.8} Pre-Registered Out-of-Sample Validation} { 11} { subsection.3.8} \protected @file@percent }
\newlabel { sec:prereg} { { 3.8} { 11} { Pre-Registered Out-of-Sample Validation} { subsection.3.8} { } }
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { paragraph} { Confirmatory versus exploratory analyses.} { 11} { section*.4} \protected @file@percent }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 3.9} Combined Timeseries: Sinusoidal Envelope Fit} { 11} { subsection.3.9} \protected @file@percent }
\newlabel { sec:sinusoid} { { 3.9} { 11} { Combined Timeseries: Sinusoidal Envelope Fit} { subsection.3.9} { } }
\newlabel { eq:modA} { { 12} { 11} { Combined Timeseries: Sinusoidal Envelope Fit} { equation.3.12} { } }
\newlabel { eq:modB} { { 13} { 11} { Combined Timeseries: Sinusoidal Envelope Fit} { equation.3.13} { } }
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ lot} { \contentsline { table} { \numberline { 1} { \ignorespaces Analysis scope and pre-registration status. Pre-specified parameters locked before data access: $ \lambda = 4 . 54 \times 10 ^ { 10 } $ , lag range $ [ - 200 , + 200 ] $ \nobreakspace { } bins, $ M \geq 4 . 5 $ , min.\ 3 stations per bin, 5-day bin size, IAAFT surrogate count $ S = 10 ^ 4 $ .\relax } } { 12} { table.caption.5} \protected @file@percent }
\providecommand * \caption @xref[2]{ \@ setref\relax \@ undefined{ #1} }
\newlabel { tab:prereg_ scope} { { 1} { 12} { Analysis scope and pre-registration status. Pre-specified parameters locked before data access: $ \lambda = 4 . 54 \times 10 ^ { 10 } $ , lag range $ [ - 200 , + 200 ] $ ~bins, $ M \geq 4 . 5 $ , min.\ 3 stations per bin, 5-day bin size, IAAFT surrogate count $ S = 10 ^ 4 $ .\relax } { table.caption.5} { } }
\newlabel { eq:bic} { { 14} { 12} { Combined Timeseries: Sinusoidal Envelope Fit} { equation.3.14} { } }
\newlabel { eq:bf} { { 15} { 12} { Combined Timeseries: Sinusoidal Envelope Fit} { equation.3.15} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 4} Results} { 12} { section.4} \protected @file@percent }
\newlabel { sec:results} { { 4} { 12} { Results} { section.4} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.1} Raw Pairwise Correlations Between CR, Seismic, and Sunspot Data} { 12} { subsection.4.1} \protected @file@percent }
\newlabel { sec:res:rawcorr} { { 4.1} { 12} { Raw Pairwise Correlations Between CR, Seismic, and Sunspot Data} { subsection.4.1} { } }
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 4.1.1} CR index: station distribution} { 13} { subsubsection.4.1.1} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 4.1.2} Seismic energy metric} { 13} { subsubsection.4.1.2} \protected @file@percent }
\newlabel { eq:energy_ bin} { { 16} { 13} { Seismic energy metric} { equation.4.16} { } }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 4.1.3} Sunspot number} { 13} { subsubsection.4.1.3} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 4.1.4} Correlation results} { 13} { subsubsection.4.1.4} \protected @file@percent }
\@ writefile{ toc} { \contentsline { paragraph} { CR vs.\ seismicity.} { 13} { section*.6} \protected @file@percent }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { Potgieter2013}
\citation { Odintsov2006}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { paragraph} { CR vs.\ sunspot number.} { 14} { section*.7} \protected @file@percent }
\@ writefile{ toc} { \contentsline { paragraph} { Sunspot vs.\ seismicity.} { 14} { section*.8} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsubsection} { \numberline { 4.1.5} Interpretation: a confounding triangle} { 14} { subsubsection.4.1.5} \protected @file@percent }
\@ writefile{ lot} { \contentsline { table} { \numberline { 2} { \ignorespaces Raw pairwise correlation statistics across three time windows. Bonferroni correction applied for $ 3 \times 3 = 9 $ tests. CR uses the per-bin station-median index ($ \hat { x } ^ { ( 50 ) } $ ). Seismic energy is $ \log _ { 10 } \! \left ( \DOTSB \sum @ \slimits @ 10 ^ { 1 . 5 M _ W } \right ) $ . Sunspot is the 365-day smoothed daily count. $ ^ { * } p _ \text { Bonf } < 0 . 05 $ , $ ^ { * * } p _ \text { Bonf } < 0 . 01 $ , $ ^ { * * * } p _ \text { Bonf } < 0 . 001 $ .\relax } } { 15} { table.caption.9} \protected @file@percent }
\newlabel { tab:rawcorr} { { 2} { 15} { Raw pairwise correlation statistics across three time windows. Bonferroni correction applied for $ 3 \times 3 = 9 $ tests. CR uses the per-bin station-median index ($ \hat { x } ^ { ( 50 ) } $ ). Seismic energy is $ \log _ { 10 } \! \left ( \sum 10 ^ { 1 . 5 M _ W } \right ) $ . Sunspot is the 365-day smoothed daily count. $ ^ { * } p _ \text { Bonf } < 0 . 05 $ , $ ^ { * * } p _ \text { Bonf } < 0 . 01 $ , $ ^ { * * * } p _ \text { Bonf } < 0 . 001 $ .\relax } { table.caption.9} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 1} { \ignorespaces Raw pairwise scatter plots for the in-sample window (1976--2019, $ N = 3 { , } 215 $ five-day bins). \textbf { Left} : CR station-median index vs $ \log _ { 10 } $ seismic energy; horizontal error bars span the station $ [ \hat { x } ^ { ( 5 ) } , \hat { x } ^ { ( 95 ) } ] $ spread. \textbf { Centre} : CR index vs 365-day smoothed sunspot number; the strong anti-correlation ($ r = - 0 . 82 $ ) reflects the Forbush decrease mechanism. \textbf { Right} : Smoothed sunspot number vs $ \log _ { 10 } $ seismic energy; thin horizontal error bars show the daily sunspot spread within each 5-day bin. All points are coloured by decimal year (plasma colormap), revealing the solar-cycle drift: successive cycles trace the same anti-correlated arc in the centre panel. Black curves are LOWESS trend lines ($ f = 0 . 4 $ ).\relax } } { 15} { figure.caption.10} \protected @file@percent }
\newlabel { fig:rawinsample} { { 1} { 15} { Raw pairwise scatter plots for the in-sample window (1976--2019, $ N = 3 { , } 215 $ five-day bins). \textbf { Left} : CR station-median index vs $ \log _ { 10 } $ seismic energy; horizontal error bars span the station $ [ \hat { x } ^ { ( 5 ) } , \hat { x } ^ { ( 95 ) } ] $ spread. \textbf { Centre} : CR index vs 365-day smoothed sunspot number; the strong anti-correlation ($ r = - 0 . 82 $ ) reflects the Forbush decrease mechanism. \textbf { Right} : Smoothed sunspot number vs $ \log _ { 10 } $ seismic energy; thin horizontal error bars show the daily sunspot spread within each 5-day bin. All points are coloured by decimal year (plasma colormap), revealing the solar-cycle drift: successive cycles trace the same anti-correlated arc in the centre panel. Black curves are LOWESS trend lines ($ f = 0 . 4 $ ).\relax } { figure.caption.10} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 2} { \ignorespaces Raw pairwise scatter plots for the out-of-sample window (2020--2025, $ N = 390 $ bins, 27 NMDB stations). Layout identical to Figure\nobreakspace { } \ref { fig:rawinsample} . The CR--sunspot anti-correlation strengthens to $ r = - 0 . 939 $ during Solar Cycle\nobreakspace { } 25, which had a particularly wide dynamic range. The CR--seismicity correlation ($ r = 0 . 046 $ , $ p _ \text { Bonf } = 1 . 0 $ ) is indistinguishable from zero.\relax } } { 16} { figure.caption.11} \protected @file@percent }
\newlabel { fig:rawoos} { { 2} { 16} { Raw pairwise scatter plots for the out-of-sample window (2020--2025, $ N = 390 $ bins, 27 NMDB stations). Layout identical to Figure~\ref { fig:rawinsample} . The CR--sunspot anti-correlation strengthens to $ r = - 0 . 939 $ during Solar Cycle~25, which had a particularly wide dynamic range. The CR--seismicity correlation ($ r = 0 . 046 $ , $ p _ \text { Bonf } = 1 . 0 $ ) is indistinguishable from zero.\relax } { figure.caption.11} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 3} { \ignorespaces Raw pairwise scatter plots for the full combined window (1976--2025, $ N = 3 { , } 604 $ bins). The multi-decadal colour gradient in the centre panel shows CR and sunspot oscillating in anti-phase across four complete solar cycles. The overall CR--seismicity Pearson correlation ($ r = 0 . 055 $ , $ \rho = 0 . 065 $ ) is statistically significant ($ p _ \text { Bonf } = 0 . 008 $ ) but quantitatively negligible, and its sign follows directly from the confounding triangle described in the text.\relax } } { 16} { figure.caption.12} \protected @file@percent }
\newlabel { fig:rawcombined} { { 3} { 16} { Raw pairwise scatter plots for the full combined window (1976--2025, $ N = 3 { , } 604 $ bins). The multi-decadal colour gradient in the centre panel shows CR and sunspot oscillating in anti-phase across four complete solar cycles. The overall CR--seismicity Pearson correlation ($ r = 0 . 055 $ , $ \rho = 0 . 065 $ ) is statistically significant ($ p _ \text { Bonf } = 0 . 008 $ ) but quantitatively negligible, and its sign follows directly from the confounding triangle described in the text.\relax } { figure.caption.12} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.2} In-Sample Replication (1976--2019)} { 17} { subsection.4.2} \protected @file@percent }
\newlabel { sec:res:insample} { { 4.2} { 17} { In-Sample Replication (1976--2019)} { subsection.4.2} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 4} { \ignorespaces Cross-correlation function $ r ( \tau ) $ for the raw (undetrended) CR index and global seismic metric, 1976--2019. The dominant peak at $ \tau = - 525 $ \nobreakspace { } days (vertical dashed line, red) corresponds to a half-solar-cycle lag; the claimed $ \tau = + 15 $ \nobreakspace { } days is marked with a vertical solid line (blue). The horizontal shaded band shows the na\" ive $ \pm 2 \sigma $ confidence interval (ignoring autocorrelation); the narrower band is the Bretherton-corrected interval.\relax } } { 17} { figure.caption.13} \protected @file@percent }
\newlabel { fig:homola} { { 4} { 17} { Cross-correlation function $ r ( \tau ) $ for the raw (undetrended) CR index and global seismic metric, 1976--2019. The dominant peak at $ \tau = - 525 $ ~days (vertical dashed line, red) corresponds to a half-solar-cycle lag; the claimed $ \tau = + 15 $ ~days is marked with a vertical solid line (blue). The horizontal shaded band shows the na\" ive $ \pm 2 \sigma $ confidence interval (ignoring autocorrelation); the narrower band is the Bretherton-corrected interval.\relax } { figure.caption.13} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.3} IAAFT Surrogate Test} { 18} { subsection.4.3} \protected @file@percent }
\newlabel { sec:res:surr} { { 4.3} { 18} { IAAFT Surrogate Test} { subsection.4.3} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 5} { \ignorespaces Null distribution of the peak cross-correlation statistic from 10{ ,} 000 IAAFT surrogates for the raw (blue) and HP-detrended (orange) CR--seismic series. Vertical dashed lines mark the observed peak for each case. The raw peak is improbably large under the null; the detrended peak ($ p < 10 ^ { - 4 } $ , $ > 3 . 9 \sigma $ ) is nominally significant but sensitive to the choice of $ N _ \text { eff } $ estimator (see Table\nobreakspace { } \ref { tab:neff} ), and occurs at a lag inconsistent with the claimed mechanism. The correlation at the claimed $ \tau = + 15 $ \nobreakspace { } d is not significant.\relax } } { 18} { figure.caption.14} \protected @file@percent }
\newlabel { fig:stress} { { 5} { 18} { Null distribution of the peak cross-correlation statistic from 10{ ,} 000 IAAFT surrogates for the raw (blue) and HP-detrended (orange) CR--seismic series. Vertical dashed lines mark the observed peak for each case. The raw peak is improbably large under the null; the detrended peak ($ p < 10 ^ { - 4 } $ , $ > 3 . 9 \sigma $ ) is nominally significant but sensitive to the choice of $ N _ \text { eff } $ estimator (see Table~\ref { tab:neff} ), and occurs at a lag inconsistent with the claimed mechanism. The correlation at the claimed $ \tau = + 15 $ ~d is not significant.\relax } { figure.caption.14} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.4} Effect of Solar-Cycle Detrending} { 19} { subsection.4.4} \protected @file@percent }
\newlabel { sec:res:detrend} { { 4.4} { 19} { Effect of Solar-Cycle Detrending} { subsection.4.4} { } }
\@ writefile{ lot} { \contentsline { table} { \numberline { 3} { \ignorespaces Cross-correlation statistics at $ \tau = + 15 $ \nobreakspace { } days under four preprocessing conditions, in-sample window 1976--2019.\relax } } { 19} { table.caption.15} \protected @file@percent }
\newlabel { tab:detrend} { { 3} { 19} { Cross-correlation statistics at $ \tau = + 15 $ ~days under four preprocessing conditions, in-sample window 1976--2019.\relax } { table.caption.15} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.5} Detrending Robustness} { 19} { subsection.4.5} \protected @file@percent }
\newlabel { sec:detrend_ robust} { { 4.5} { 19} { Detrending Robustness} { subsection.4.5} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.6} Comparison of $ N _ \text { eff } $ Estimators} { 19} { subsection.4.6} \protected @file@percent }
\newlabel { sec:neff_ comparison} { { 4.6} { 19} { Comparison of $ N _ \text { eff } $ Estimators} { subsection.4.6} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 6} { \ignorespaces Cross-correlation functions for the raw (blue) and HP-detrended (orange) series. The dominant peak at $ \tau = - 525 $ \nobreakspace { } days in the raw data (dashed blue) is absent after detrending, confirming it is a solar-cycle artefact. Neither series exhibits a significant peak at $ \tau = + 15 $ \nobreakspace { } days (vertical grey line).\relax } } { 20} { figure.caption.16} \protected @file@percent }
\newlabel { fig:detrended} { { 6} { 20} { Cross-correlation functions for the raw (blue) and HP-detrended (orange) series. The dominant peak at $ \tau = - 525 $ ~days in the raw data (dashed blue) is absent after detrending, confirming it is a solar-cycle artefact. Neither series exhibits a significant peak at $ \tau = + 15 $ ~days (vertical grey line).\relax } { figure.caption.16} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 7} { \ignorespaces Cross-correlation $ r ( \tau ) $ under three detrending approaches (HP filter, Butterworth highpass, rolling-mean subtraction) for the raw CR index vs.\ log$ _ { 10 } $ seismic energy, 1976--2019. The vertical grey line marks the claimed $ \tau = + 15 $ \nobreakspace { } days. No method produces a positive feature at this lag.\relax } } { 20} { figure.caption.17} \protected @file@percent }
\newlabel { fig:detrend_ robust} { { 7} { 20} { Cross-correlation $ r ( \tau ) $ under three detrending approaches (HP filter, Butterworth highpass, rolling-mean subtraction) for the raw CR index vs.\ log$ _ { 10 } $ seismic energy, 1976--2019. The vertical grey line marks the claimed $ \tau = + 15 $ ~days. No method produces a positive feature at this lag.\relax } { figure.caption.17} { } }
\@ writefile{ lot} { \contentsline { table} { \numberline { 4} { \ignorespaces Comparison of $ N _ \text { eff } $ estimators for raw in-sample series at $ \tau = + 15 $ \nobreakspace { } days ($ N = 3 { , } 214 $ , $ r = 0 . 079 $ ).\relax } } { 21} { table.caption.18} \protected @file@percent }
\newlabel { tab:neff} { { 4} { 21} { Comparison of $ N _ \text { eff } $ estimators for raw in-sample series at $ \tau = + 15 $ ~days ($ N = 3 { , } 214 $ , $ r = 0 . 079 $ ).\relax } { table.caption.18} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.7} Magnitude Threshold Sensitivity} { 21} { subsection.4.7} \protected @file@percent }
\newlabel { sec:mag_ threshold} { { 4.7} { 21} { Magnitude Threshold Sensitivity} { subsection.4.7} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 8} { \ignorespaces Cross-correlation $ r ( \tau ) $ for three magnitude cutoffs ($ M \geq 4 . 5 $ , blue; $ M \geq 5 . 0 $ , orange; $ M \geq 6 . 0 $ , green), in-sample window 1976--2019. The vertical grey line marks $ \tau = + 15 $ \nobreakspace { } days. All three cutoffs yield similar, small correlations at the claimed lag, and the dominant peak location ($ \tau = - 525 $ \nobreakspace { } days) is stable.\relax } } { 21} { figure.caption.19} \protected @file@percent }
\newlabel { fig:mag_ threshold} { { 8} { 21} { Cross-correlation $ r ( \tau ) $ for three magnitude cutoffs ($ M \geq 4 . 5 $ , blue; $ M \geq 5 . 0 $ , orange; $ M \geq 6 . 0 $ , green), in-sample window 1976--2019. The vertical grey line marks $ \tau = + 15 $ ~days. All three cutoffs yield similar, small correlations at the claimed lag, and the dominant peak location ($ \tau = - 525 $ ~days) is stable.\relax } { figure.caption.19} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.8} Block-Bootstrap Surrogate Comparison} { 21} { subsection.4.8} \protected @file@percent }
\newlabel { sec:res:blockbootstrap} { { 4.8} { 21} { Block-Bootstrap Surrogate Comparison} { subsection.4.8} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 9} { \ignorespaces Block-bootstrap null distributions for the raw CR--seismic pair ($ B = 804 $ bins $ \approx 11 $ \nobreakspace { } yr, $ S = 5 { , } 000 $ surrogates). \textbf { Left} : distribution of $ r ( + 15 \, \text { d } ) $ under the CBB null; observed value $ r = 0 . 079 $ (red) lies at the $ p = 0 . 022 $ tail. \textbf { Right} : distribution of the peak $ |r| $ ; observed peak $ 0 . 135 $ at $ \tau = - 525 $ \nobreakspace { } d lies at $ p = 0 . 008 $ . Grey dashed lines mark the 95\% CI of the null.\relax } } { 22} { figure.caption.20} \protected @file@percent }
\newlabel { fig:blockbootstrap} { { 9} { 22} { Block-bootstrap null distributions for the raw CR--seismic pair ($ B = 804 $ bins $ \approx 11 $ ~yr, $ S = 5 { , } 000 $ surrogates). \textbf { Left} : distribution of $ r ( + 15 \, \text { d } ) $ under the CBB null; observed value $ r = 0 . 079 $ (red) lies at the $ p = 0 . 022 $ tail. \textbf { Right} : distribution of the peak $ |r| $ ; observed peak $ 0 . 135 $ at $ \tau = - 525 $ ~d lies at $ p = 0 . 008 $ . Grey dashed lines mark the 95\% CI of the null.\relax } { figure.caption.20} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.9} Partial Correlation: Controlling for Sunspot Number} { 22} { subsection.4.9} \protected @file@percent }
\newlabel { sec:res:partialcorr} { { 4.9} { 22} { Partial Correlation: Controlling for Sunspot Number} { subsection.4.9} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.10} Spectral Coherence and Mutual Information} { 22} { subsection.4.10} \protected @file@percent }
\newlabel { sec:res:coherence} { { 4.10} { 22} { Spectral Coherence and Mutual Information} { subsection.4.10} { } }
\@ writefile{ toc} { \contentsline { paragraph} { Coherence.} { 22} { section*.22} \protected @file@percent }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 10} { \ignorespaces Cross-correlation $ r ( \tau ) $ for the raw seismic metric (blue) and the sunspot-regressed seismic residual (orange), both against the CR index, on the raw (unfiltered) in-sample series. The claimed $ \tau = + 15 $ \nobreakspace { } d (grey dashed line) shows $ r _ \text { raw } = 0 . 079 $ vs.\ $ r _ \text { partial } = 0 . 029 $ , a 63\% reduction once the shared solar-cycle component is regressed out.\relax } } { 23} { figure.caption.21} \protected @file@percent }
\newlabel { fig:partialcorr} { { 10} { 23} { Cross-correlation $ r ( \tau ) $ for the raw seismic metric (blue) and the sunspot-regressed seismic residual (orange), both against the CR index, on the raw (unfiltered) in-sample series. The claimed $ \tau = + 15 $ ~d (grey dashed line) shows $ r _ \text { raw } = 0 . 079 $ vs.\ $ r _ \text { partial } = 0 . 029 $ , a 63\% reduction once the shared solar-cycle component is regressed out.\relax } { figure.caption.21} { } }
\@ writefile{ toc} { \contentsline { paragraph} { Mutual information.} { 23} { section*.23} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.11} Missing-Data Sensitivity} { 23} { subsection.4.11} \protected @file@percent }
\newlabel { sec:res:missing} { { 4.11} { 23} { Missing-Data Sensitivity} { subsection.4.11} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 11} { \ignorespaces \textbf { Left} : Magnitude-squared coherence between the CR index and seismic metric (blue), with the solar-cycle band shaded (orange, 0.08--0.115 cycles\nobreakspace { } yr$ ^ { - 1 } $ ) and the 95\% significance level (red dashed). The mean coherence in the SC band is 0.840, confirming a strong shared solar-cycle component. \textbf { Right} : kNN mutual information ($ k = 5 $ ) at lag $ \tau = 0 $ (blue) and $ \tau = + 15 $ \nobreakspace { } d (orange) vs.\ their respective shuffle-null distributions. Both observed MI values are indistinguishable from zero; $ p ( + 15 \, \text { d } ) = 1 . 000 $ .\relax } } { 24} { figure.caption.24} \protected @file@percent }
\newlabel { fig:coherence} { { 11} { 24} { \textbf { Left} : Magnitude-squared coherence between the CR index and seismic metric (blue), with the solar-cycle band shaded (orange, 0.08--0.115 cycles~yr$ ^ { - 1 } $ ) and the 95\% significance level (red dashed). The mean coherence in the SC band is 0.840, confirming a strong shared solar-cycle component. \textbf { Right} : kNN mutual information ($ k = 5 $ ) at lag $ \tau = 0 $ (blue) and $ \tau = + 15 $ ~d (orange) vs.\ their respective shuffle-null distributions. Both observed MI values are indistinguishable from zero; $ p ( + 15 \, \text { d } ) = 1 . 000 $ .\relax } { figure.caption.24} { } }
\@ writefile{ lot} { \contentsline { table} { \numberline { 5} { \ignorespaces Missing-data sensitivity: global CR index and correlation at $ \tau = + 15 $ \nobreakspace { } days for three station-threshold values.\relax } } { 24} { table.caption.25} \protected @file@percent }
\newlabel { tab:missing} { { 5} { 24} { Missing-data sensitivity: global CR index and correlation at $ \tau = + 15 $ ~days for three station-threshold values.\relax } { table.caption.25} { } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { GardnerKnopoff1974}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.12} Bin-Size Sensitivity} { 25} { subsection.4.12} \protected @file@percent }
\newlabel { sec:res:binsize} { { 4.12} { 25} { Bin-Size Sensitivity} { subsection.4.12} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 12} { \ignorespaces Cross-correlation $ r ( \tau ) $ for three bin sizes: 1-day (left), 5-day (centre), 27-day (right). The dominant peak (red dotted) consistently falls near $ \tau \approx - 520 $ \nobreakspace { } days across all bin sizes. The correlation at the claimed $ \tau = + 15 $ \nobreakspace { } days (grey dashed) increases from 0.036 to 0.123 as bin size increases, consistent with increasing solar-cycle leakage rather than a physical short-lag signal.\relax } } { 25} { figure.caption.26} \protected @file@percent }
\newlabel { fig:binsize} { { 12} { 25} { Cross-correlation $ r ( \tau ) $ for three bin sizes: 1-day (left), 5-day (centre), 27-day (right). The dominant peak (red dotted) consistently falls near $ \tau \approx - 520 $ ~days across all bin sizes. The correlation at the claimed $ \tau = + 15 $ ~days (grey dashed) increases from 0.036 to 0.123 as bin size increases, consistent with increasing solar-cycle leakage rather than a physical short-lag signal.\relax } { figure.caption.26} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.13} Earthquake Declustering (Gardner--Knopoff)} { 25} { subsection.4.13} \protected @file@percent }
\newlabel { sec:res:decluster} { { 4.13} { 25} { Earthquake Declustering (Gardner--Knopoff)} { subsection.4.13} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 13} { \ignorespaces Cross-correlation for the full catalogue ($ n = 232 { , } 043 $ events, blue) and the Gardner--Knopoff declustered catalogue ($ n = 166 { , } 169 $ mainshocks, orange), in-sample 1976--2019. Removing 28.4\% of events as aftershocks changes $ r ( + 15 \, \text { d } ) $ by only $ \Delta r = 0 . 014 $ , confirming the result is not driven by aftershock swarms.\relax } } { 26} { figure.caption.27} \protected @file@percent }
\newlabel { fig:decluster} { { 13} { 26} { Cross-correlation for the full catalogue ($ n = 232 { , } 043 $ events, blue) and the Gardner--Knopoff declustered catalogue ($ n = 166 { , } 169 $ mainshocks, orange), in-sample 1976--2019. Removing 28.4\% of events as aftershocks changes $ r ( + 15 \, \text { d } ) $ by only $ \Delta r = 0 . 014 $ , confirming the result is not driven by aftershock swarms.\relax } { figure.caption.27} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.14} Sub-Period Analysis by Solar Cycle} { 26} { subsection.4.14} \protected @file@percent }
\newlabel { sec:res:subcycles} { { 4.14} { 26} { Sub-Period Analysis by Solar Cycle} { subsection.4.14} { } }
\@ writefile{ lot} { \contentsline { table} { \numberline { 6} { \ignorespaces Per-solar-cycle cross-correlation at $ \tau = + 15 $ \nobreakspace { } days and the within-cycle dominant peak.\relax } } { 26} { table.caption.28} \protected @file@percent }
\newlabel { tab:subcycles} { { 6} { 26} { Per-solar-cycle cross-correlation at $ \tau = + 15 $ ~days and the within-cycle dominant peak.\relax } { table.caption.28} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.15} Geographic Localisation} { 26} { subsection.4.15} \protected @file@percent }
\newlabel { sec:res:geo} { { 4.15} { 26} { Geographic Localisation} { subsection.4.15} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 14} { \ignorespaces Cross-correlation $ r ( \tau ) $ within each of the four complete solar cycles (21--24) of the in-sample period (1976--2019). The claimed lag $ \tau = + 15 $ \nobreakspace { } days (grey dashed) shows modest positive correlations (0.018--0.073), but the dominant peak lag (red dotted) is inconsistent across cycles ($ - 65 $ , $ - 125 $ , $ + 125 $ , $ - 125 $ \nobreakspace { } days), pointing to a drift in the relative solar-cycle phase rather than a physical mechanism.\relax } } { 27} { figure.caption.29} \protected @file@percent }
\newlabel { fig:subcycles} { { 14} { 27} { Cross-correlation $ r ( \tau ) $ within each of the four complete solar cycles (21--24) of the in-sample period (1976--2019). The claimed lag $ \tau = + 15 $ ~days (grey dashed) shows modest positive correlations (0.018--0.073), but the dominant peak lag (red dotted) is inconsistent across cycles ($ - 65 $ , $ - 125 $ , $ + 125 $ , $ - 125 $ ~days), pointing to a drift in the relative solar-cycle phase rather than a physical mechanism.\relax } { figure.caption.29} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 15} { \ignorespaces Heatmap of BH-significant station--grid-cell pairs ($ q = 0 . 05 $ ). Each row is an NMDB station; each column is a $ 10 ° \times 10 ° $ seismic grid cell. Significant pairs (455/7{ ,} 037) are scattered without obvious geographic clustering, inconsistent with a local coupling mechanism.\relax } } { 28} { figure.caption.30} \protected @file@percent }
\newlabel { fig:geoheatmap} { { 15} { 28} { Heatmap of BH-significant station--grid-cell pairs ($ q = 0 . 05 $ ). Each row is an NMDB station; each column is a $ 10 ° \times 10 ° $ seismic grid cell. Significant pairs (455/7{ ,} 037) are scattered without obvious geographic clustering, inconsistent with a local coupling mechanism.\relax } { figure.caption.30} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 16} { \ignorespaces Optimal lag $ \tau ^ * ( s,g ) $ vs.\ great-circle distance $ d ( s,g ) $ for all 7{ ,} 037 station--cell pairs (grey) and BH-significant pairs (coloured by peak $ |r| $ ). The OLS regression line (red) has slope $ \beta = - 0 . 45 $ \nobreakspace { } days/1000\, km ($ p = 0 . 21 $ ), indistinguishable from zero. A local propagation or diffusion mechanism predicts a positive slope; the null result is inconsistent with such models but does not exclude globally instantaneous coupling.\relax } } { 28} { figure.caption.31} \protected @file@percent }
\newlabel { fig:geodistlag} { { 16} { 28} { Optimal lag $ \tau ^ * ( s,g ) $ vs.\ great-circle distance $ d ( s,g ) $ for all 7{ ,} 037 station--cell pairs (grey) and BH-significant pairs (coloured by peak $ |r| $ ). The OLS regression line (red) has slope $ \beta = - 0 . 45 $ ~days/1000\, km ($ p = 0 . 21 $ ), indistinguishable from zero. A local propagation or diffusion mechanism predicts a positive slope; the null result is inconsistent with such models but does not exclude globally instantaneous coupling.\relax } { figure.caption.31} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.16} Pre-Registered Out-of-Sample Validation (2020--2025)} { 29} { subsection.4.16} \protected @file@percent }
\newlabel { sec:res:oos} { { 4.16} { 29} { Pre-Registered Out-of-Sample Validation (2020--2025)} { subsection.4.16} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 17} { \ignorespaces Out-of-sample cross-correlation function (2020--2025, $ T = 390 $ bins, $ 10 ^ 5 $ phase surrogates). The observed $ r ( \tau ) $ (black) lies within the surrogate 95th-percentile envelope (grey shading). The claimed signal at $ \tau = + 15 $ \nobreakspace { } d (vertical line) is $ r = 0 . 030 $ --- below the surrogate 95th percentile of 0.101.\relax } } { 29} { figure.caption.32} \protected @file@percent }
\newlabel { fig:oosxcorr} { { 17} { 29} { Out-of-sample cross-correlation function (2020--2025, $ T = 390 $ bins, $ 10 ^ 5 $ phase surrogates). The observed $ r ( \tau ) $ (black) lies within the surrogate 95th-percentile envelope (grey shading). The claimed signal at $ \tau = + 15 $ ~d (vertical line) is $ r = 0 . 030 $ --- below the surrogate 95th percentile of 0.101.\relax } { figure.caption.32} { } }
\@ writefile{ lot} { \contentsline { table} { \numberline { 7} { \ignorespaces Pre-registered prediction scorecard for the out-of-sample window.\relax } } { 29} { table.caption.33} \protected @file@percent }
\newlabel { tab:prereg} { { 7} { 29} { Pre-registered prediction scorecard for the out-of-sample window.\relax } { table.caption.33} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 18} { \ignorespaces Rolling $ r ( + 15 \, \text { d } ) $ in 18-month overlapping windows across the out-of-sample period. Error bars are bootstrap 95\% confidence intervals. The grey horizontal band shows the surrogate 95th percentile. The signal shows no consistent sign or trend.\relax } } { 30} { figure.caption.34} \protected @file@percent }
\newlabel { fig:rolling} { { 18} { 30} { Rolling $ r ( + 15 \, \text { d } ) $ in 18-month overlapping windows across the out-of-sample period. Error bars are bootstrap 95\% confidence intervals. The grey horizontal band shows the surrogate 95th percentile. The signal shows no consistent sign or trend.\relax } { figure.caption.34} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 4.17} Combined 1976--2025 Analysis: Sinusoidal Modulation} { 30} { subsection.4.17} \protected @file@percent }
\newlabel { sec:res:combined} { { 4.17} { 30} { Combined 1976--2025 Analysis: Sinusoidal Modulation} { subsection.4.17} { } }
\@ writefile{ lof} { \contentsline { figure} { \numberline { 19} { \ignorespaces Annual rolling $ r ( + 15 \, \text { d } ) $ across the full 1976--2025 period (grey points with 95\% bootstrap CI). The sinusoidal best-fit (red curve, $ P = 13 . 0 $ \nobreakspace { } yr) and the constant-mean model (dashed) are nearly indistinguishable; BIC comparison slightly favours the constant model ($ \text { BF } = 0 . 75 $ ). The vertical dashed line marks the in-sample/out-of-sample split (2020).\relax } } { 30} { figure.caption.35} \protected @file@percent }
\newlabel { fig:combined} { { 19} { 30} { Annual rolling $ r ( + 15 \, \text { d } ) $ across the full 1976--2025 period (grey points with 95\% bootstrap CI). The sinusoidal best-fit (red curve, $ P = 13 . 0 $ ~yr) and the constant-mean model (dashed) are nearly indistinguishable; BIC comparison slightly favours the constant model ($ \text { BF } = 0 . 75 $ ). The vertical dashed line marks the in-sample/out-of-sample split (2020).\relax } { figure.caption.35} { } }
\citation { KassRaftery1995}
2026-04-24 05:51:30 +00:00
\citation { Homola2023}
\citation { Odintsov2006}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 5} Discussion} { 31} { section.5} \protected @file@percent }
\newlabel { sec:discussion} { { 5} { 31} { Discussion} { section.5} { } }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 5.1} Why Does the Raw Correlation Appear So Strong?} { 31} { subsection.5.1} \protected @file@percent }
2026-04-24 05:51:30 +00:00
\citation { Aplin2005}
\citation { Pulinets2004}
\citation { Homola2023}
\citation { Urata2018}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 5.2} Physical Plausibility of the Claimed Mechanism} { 32} { subsection.5.2} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 5.3} Comparison with Prior Replication Attempts} { 32} { subsection.5.3} \protected @file@percent }
\@ writefile{ toc} { \contentsline { subsection} { \numberline { 5.4} Limitations} { 32} { subsection.5.4} \protected @file@percent }
2026-04-24 05:51:30 +00:00
\citation { Homola2023}
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\citation { Homola2023}
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\@ writefile{ toc} { \contentsline { section} { \numberline { 6} Conclusions} { 33} { section.6} \protected @file@percent }
\newlabel { sec:conclusions} { { 6} { 33} { Conclusions} { section.6} { } }
2026-04-24 05:51:30 +00:00
\bibstyle { plainnat}
\bibdata { refs}
\bibcite { Aplin2005} { { 1} { 2006} { { Aplin} } { { } } }
Add seven additional robustness checks (3a–3g) and update paper
3a Block-bootstrap surrogates (B=804 bins≈11 yr, n=5000): raw r(+15d)
p=0.022 — marginally significant on undetrended series, driven by
the shared solar-cycle trend not a causal signal.
3b Partial correlation (sunspot-regressed, no filter): r(+15d) drops
from 0.079 to 0.029 (63%) — confirms solar-cycle confounding without
preprocessing circularity.
3c Spectral coherence in solar-cycle band = 0.840 (>0.776 threshold);
kNN mutual information at τ=+15d = 0.000 nats (p=1.000) — no
nonlinear dependence at the claimed lag.
3d Missing-data impact: 0% NaN at station thresholds 2/3/5; r(+15d)
unchanged — missing data is not a confound.
3e Bin-size sensitivity: dominant peak at τ≈-520d for 1/5/27-day bins;
r at +15d scales with bin size (solar-cycle leakage, not physical).
3f GK declustering removes 28.4% of events as aftershocks; r(+15d)
changes by only Δ=0.014 — aftershock clustering not a confound.
3g Per-solar-cycle analysis (cycles 21–24): r(+15d) all positive
(0.018–0.073) but dominant peak lags scattered (-65/-125/+125/-125d)
— phase-drifting solar-cycle artefact, not physical precursor.
Script fixes: vectorised block bootstrap (401×5000 → NumPy matmul),
kNN MI with sorted searchsorted (O(N log N), no inflated values),
coherence nperseg 512→2048 (resolves solar-cycle band), axhspan→axvspan.
Paper: new Methods subsections (block bootstrap, partial correlation,
nonlinear dependence); new Results subsections for each check; updated
Conclusions with 7-item robustness summary; Kraskov2004 and
GardnerKnopoff1974 added to refs.bib; Homola2023 updated to arXiv
preprint 2204.12310; eq:energy duplicate label fixed. PDF: 35 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 18:43:08 +00:00
\bibcite { Bartlett1946} { { 2} { 1946} { { Bartlett} } { { } } }
\bibcite { Benjamini1995} { { 3} { 1995} { { Benjamini and Hochberg} } { { } } }
\bibcite { Bretherton1999} { { 4} { 1999} { { Bretherton et~al.} } { { Bretherton, Widmann, Dymnikov, Wallace, and Blade} } }
\bibcite { Cleveland1990} { { 5} { 1990} { { Cleveland et~al.} } { { Cleveland, Cleveland, McRae, and Terpenning} } }
\bibcite { GardnerKnopoff1974} { { 6} { 1974} { { Gardner and Knopoff} } { { } } }
\bibcite { HP1997} { { 7} { 1997} { { Hodrick and Prescott} } { { } } }
\bibcite { Homola2023} { { 8} { 2022} { { Homola et~al.} } { { } } }
\bibcite { Kanamori1977} { { 9} { 1977} { { Kanamori} } { { } } }
Editorial revision: precision, framing, and internal consistency (4a–4k)
4a Moderate principal conclusion: 'no statistically credible evidence'
replaced with 'no statistically robust evidence within tested frameworks'
plus clause acknowledging untested mechanisms (threshold effects,
nonlinear triggering, extreme-event coupling).
4b Fix independence-assumption framing: 'physically invalid seismic metric'
→ 'physically inappropriate'; separate claim about naive p-values now
reads 'statistically invalid under the violated serial-independence
assumption (autocorrelation inflates nominal sample size by 3–5×)'.
4c 3.9σ detrended peak no longer called 'marginal': figure caption and
nearby text now read 'nominally significant but sensitive to Neff
estimation, at a lag inconsistent with the claimed mechanism'.
4d CR terminology standardised: 'global CR index' defined precisely at
first use in Data section (dimensionless, station mean ≡ 1, ≥3 stations
per bin); 'CR flux' retained only for the physical quantity.
4e Geographic conclusion reframed: 'no local mechanism' replaced with
'inconsistent with simple wave-propagation or diffusion models, but does
not rule out instantaneous global coupling mechanisms (e.g. atmospheric
electric field modulation)'.
4f Bayes factor qualified: parenthetical after BF=0.75 notes the restricted
two-hypothesis model space and cites Kass & Raftery (1995).
4g OOS limitations expanded: explicit paragraph noting the 5-yr window
with no complete solar cycle, limited statistical power, and that
p_global=0.100 is consistent with—rather than strong evidence against
—the claim; OOS failure downweighted vs 44-yr in-sample analysis.
4h Confirmatory vs exploratory scope table added (Table tab:prereg_scope)
listing pre-specified parameters and which analyses were confirmatory
vs post-hoc exploratory.
4i Alternative solar-cycle confounds acknowledged in Discussion: geomagnetic
activity cycles and long-term seismic clustering added as alternative
explanations for the shared 10-year periodicity.
4j Fixed: 'Out-of-sample poos from script 08' removed from Limitations;
GitHub URL removed from abstract (kept in Data Availability only);
Discussion run-on sentences broken up.
4k Abstract rewritten to ≤250 words in five-part structure: prior claim,
data/methods (two sentences), key quantitative results, scoped
interpretation, one-sentence limitation. Causal language qualified.
Also adds KassRaftery1995 to refs.bib. PDF: 36 pages.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-24 19:23:57 +00:00
\bibcite { KassRaftery1995} { { 10} { 1995} { { Kass and Raftery} } { { } } }
\bibcite { Kraskov2004} { { 11} { 2004} { { Kraskov et~al.} } { { Kraskov, St{ \" o} gbauer, and Grassberger} } }
\bibcite { Odintsov2006} { { 12} { 2006} { { Odintsov et~al.} } { { Odintsov, Boyarchuk, Georgieva, Kirov, and Atanasov} } }
\bibcite { Potgieter2013} { { 13} { 2013} { { Potgieter} } { { } } }
\bibcite { Pulinets2004} { { 14} { 2004} { { Pulinets and Boyarchuk} } { { } } }
\bibcite { RahnUhlig2002} { { 15} { 2002} { { Ravn and Uhlig} } { { } } }
\bibcite { Schreiber2000} { { 16} { 2000} { { Schreiber and Schmitz} } { { } } }
\bibcite { SIDC2024} { { 17} { 2024} { { SILSO World Data Center} } { { } } }
\bibcite { Stoupel1990} { { 18} { 1990} { { Stoupel} } { { } } }
\bibcite { Tavares2011} { { 19} { 2011} { { Tavares and Azevedo} } { { } } }
\bibcite { Theiler1992} { { 20} { 1992} { { Theiler et~al.} } { { Theiler, Eubank, Longtin, Galdrikian, and Farmer} } }
\bibcite { Urata2018} { { 21} { 2018} { { Urata and Tanimoto} } { { } } }
\bibcite { USGS2024} { { 22} { 2024} { { USGS Earthquake Hazards Program} } { { } } }
\gdef \@ abspage@last{ 36}
