Cross-Domain Detector Meta-Analysis Report
Generated: 2026-07-16 22:44 UTC
This report is produced automatically from the committed detector-evidence aggregates under examples/real_data/*/. It normalises each detector's performance, ranks detectors within every domain, and derives a ranked backlog of refinement candidates.
Data sources
| Domain |
Source file |
Schema |
| afdb_atrial_fibrillation |
early_warning_leadtime_cardiac_results.json |
early-warning lead-time |
| afdb_atrial_fibrillation_multiscale |
early_warning_leadtime_cardiac_multiscale_results.json |
early-warning lead-time |
| cap_kuramoto_variants |
cap_kuramoto_variants.json |
early-warning lead-time |
| cap_multichannel_staging |
cap_multichannel_aggregate.json |
honest-audit aggregate |
| chb01_seizures |
early_warning_leadtime_eeg_results.json |
early-warning lead-time |
| chb01_seizures_multiscale |
early_warning_leadtime_eeg_multiscale_results.json |
early-warning lead-time |
| csd_variant_synthetic |
csd_variant_synthetic_results.json |
early-warning lead-time |
| dakos_climate_transitions |
early_warning_leadtime_climate_results.json |
early-warning lead-time |
| dakos_climate_transitions_multiscale |
early_warning_leadtime_climate_multiscale_results.json |
early-warning lead-time |
| psml_grid_oscillation |
early_warning_leadtime_grid_results.json |
early-warning lead-time |
| psml_grid_oscillation_multiscale |
early_warning_leadtime_grid_multiscale_results.json |
early-warning lead-time |
| regime_adaptive_ensemble |
regime_adaptive_ensemble.json |
early-warning lead-time |
| sleepedf_kuramoto_variants |
sleepedf_kuramoto_variants.json |
early-warning lead-time |
| synthetic_honest_audit_demo |
synthetic_honest_audit_demo.json |
honest-audit aggregate |
Per-domain rankings
BH-adj p is the Benjamini–Hochberg false-discovery-rate-adjusted p-value, corrected across the detectors compared on that domain, so a raw p-value that only looks significant because several detectors were tried is not read as a discovery. The raw p-value is kept beside it.
afdb_atrial_fibrillation
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
ensemble_weighted |
33.3% |
2.193e-01 |
4.386e-01 |
False |
| 1 |
synchronisation |
33.3% |
2.193e-01 |
4.386e-01 |
False |
| 3 |
critical_slowing_down |
16.7% |
5.627e-01 |
7.503e-01 |
False |
| 4 |
transition_entropy |
0.0% |
1.000e+00 |
1.000e+00 |
False |
afdb_atrial_fibrillation_multiscale
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down_multiscale |
33.3% |
2.193e-01 |
3.655e-01 |
False |
| 1 |
ensemble_weighted |
33.3% |
2.193e-01 |
3.655e-01 |
False |
| 1 |
synchronisation |
33.3% |
2.193e-01 |
3.655e-01 |
False |
| 4 |
critical_slowing_down |
16.7% |
5.627e-01 |
7.034e-01 |
False |
| 5 |
transition_entropy |
0.0% |
1.000e+00 |
1.000e+00 |
False |
cap_kuramoto_variants
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
coherent_sustained_kuramoto |
55.2% |
1.112e-03 |
3.054e-03 |
True |
| 2 |
normalized_delta_envelope |
52.5% |
1.000e-03 |
3.054e-03 |
True |
| 3 |
amplitude_gated_delta_kuramoto |
45.7% |
1.309e-03 |
3.054e-03 |
True |
| 4 |
adaptive_channel_kuramoto |
20.4% |
6.570e-03 |
1.150e-02 |
True |
| 5 |
multi_channel_delta_kuramoto |
18.4% |
1.486e-02 |
1.883e-02 |
True |
| 6 |
snr_weighted_delta_kuramoto |
17.5% |
1.614e-02 |
1.883e-02 |
True |
| 7 |
sustained_delta_kuramoto |
17.3% |
2.320e-02 |
2.320e-02 |
True |
cap_multichannel_staging
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
normalized_delta_envelope |
52.5% |
1.000e-03 |
3.000e-03 |
True |
| 2 |
multi_channel_delta_kuramoto |
18.4% |
1.486e-02 |
1.614e-02 |
True |
| 3 |
snr_weighted_delta_kuramoto |
17.5% |
1.614e-02 |
1.614e-02 |
True |
chb01_seizures
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down |
33.3% |
2.155e-01 |
7.465e-01 |
False |
| 2 |
ensemble_weighted |
16.7% |
5.582e-01 |
7.465e-01 |
False |
| 3 |
synchronisation |
16.7% |
5.598e-01 |
7.465e-01 |
False |
| 4 |
transition_entropy |
0.0% |
1.000e+00 |
1.000e+00 |
False |
chb01_seizures_multiscale
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down_multiscale |
33.3% |
2.154e-01 |
5.387e-01 |
False |
| 2 |
critical_slowing_down |
33.3% |
2.155e-01 |
5.387e-01 |
False |
| 3 |
ensemble_weighted |
16.7% |
5.582e-01 |
6.998e-01 |
False |
| 4 |
synchronisation |
16.7% |
5.598e-01 |
6.998e-01 |
False |
| 5 |
transition_entropy |
0.0% |
1.000e+00 |
1.000e+00 |
False |
csd_variant_synthetic
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down_multiscale |
50.0% |
4.055e-03 |
1.216e-02 |
True |
| 2 |
critical_slowing_down_surrogate |
31.2% |
2.078e-02 |
3.116e-02 |
True |
| 3 |
critical_slowing_down_baseline |
25.0% |
1.473e-01 |
1.473e-01 |
True |
dakos_climate_transitions
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down |
16.7% |
5.144e-01 |
5.144e-01 |
False |
dakos_climate_transitions_multiscale
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down_multiscale |
0.0% |
1.000e+00 |
1.000e+00 |
False |
psml_grid_oscillation
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down |
25.0% |
1.953e-01 |
5.430e-01 |
False |
| 2 |
transition_entropy |
16.7% |
4.045e-01 |
5.430e-01 |
False |
| 3 |
ensemble_weighted |
16.7% |
4.073e-01 |
5.430e-01 |
False |
| 4 |
synchronisation |
0.0% |
1.000e+00 |
1.000e+00 |
False |
psml_grid_oscillation_multiscale
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
critical_slowing_down_multiscale |
33.3% |
7.859e-02 |
3.930e-01 |
False |
| 2 |
critical_slowing_down |
25.0% |
1.953e-01 |
4.882e-01 |
False |
| 3 |
transition_entropy |
16.7% |
4.045e-01 |
5.091e-01 |
False |
| 4 |
ensemble_weighted |
16.7% |
4.073e-01 |
5.091e-01 |
False |
| 5 |
synchronisation |
0.0% |
1.000e+00 |
1.000e+00 |
False |
regime_adaptive_ensemble
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
normalized_delta_envelope |
61.9% |
6.310e-04 |
1.409e-03 |
True |
| 2 |
regime_adaptive_full |
61.5% |
7.250e-04 |
1.409e-03 |
True |
| 2 |
regime_adaptive_montage |
61.5% |
7.250e-04 |
1.409e-03 |
True |
| 4 |
coherent_sustained_kuramoto |
49.5% |
6.870e-04 |
1.409e-03 |
True |
| 5 |
amplitude_gated_delta_kuramoto |
42.1% |
7.830e-04 |
1.409e-03 |
True |
| 6 |
adaptive_channel_kuramoto |
16.3% |
1.795e-02 |
2.692e-02 |
True |
| 7 |
multi_channel_delta_kuramoto |
14.7% |
3.448e-02 |
4.145e-02 |
True |
| 8 |
snr_weighted_delta_kuramoto |
14.0% |
3.684e-02 |
4.145e-02 |
True |
| 9 |
sustained_delta_kuramoto |
13.9% |
4.926e-02 |
4.926e-02 |
True |
sleepedf_kuramoto_variants
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
normalized_delta_envelope |
99.5% |
1.000e-04 |
2.333e-04 |
True |
| 2 |
amplitude_gated_delta_kuramoto |
27.7% |
1.000e-04 |
2.333e-04 |
True |
| 3 |
coherent_sustained_kuramoto |
26.8% |
1.000e-04 |
2.333e-04 |
True |
| 4 |
adaptive_channel_kuramoto |
0.0% |
1.000e+00 |
1.000e+00 |
False |
| 4 |
multi_channel_delta_kuramoto |
0.0% |
1.000e+00 |
1.000e+00 |
False |
| 4 |
snr_weighted_delta_kuramoto |
0.0% |
1.000e+00 |
1.000e+00 |
False |
| 4 |
sustained_delta_kuramoto |
0.0% |
1.000e+00 |
1.000e+00 |
False |
synthetic_honest_audit_demo
| Rank |
Detector |
Detection rate |
p-value |
BH-adj p |
Beats chance |
| 1 |
lag1_autocorrelation |
100.0% |
1.000e-04 |
2.000e-04 |
True |
| 2 |
window_mean_control |
0.0% |
1.000e+00 |
1.000e+00 |
False |
Cross-domain overall ranking
| Rank |
Detector |
Mean rank |
Domains present |
Wins |
Domain wins |
| 1 |
critical_slowing_down_multiscale |
1.00 |
5 |
5 |
afdb_atrial_fibrillation_multiscale (1), chb01_seizures_multiscale (1), csd_variant_synthetic (1), dakos_climate_transitions_multiscale (1), psml_grid_oscillation_multiscale (1) |
| 2 |
lag1_autocorrelation |
1.00 |
1 |
1 |
synthetic_honest_audit_demo (1) |
| 3 |
normalized_delta_envelope |
1.25 |
4 |
3 |
cap_multichannel_staging (1), regime_adaptive_ensemble (1), sleepedf_kuramoto_variants (1) |
| 4 |
critical_slowing_down |
2.00 |
7 |
3 |
chb01_seizures (1), dakos_climate_transitions (1), psml_grid_oscillation (1) |
| 5 |
critical_slowing_down_surrogate |
2.00 |
1 |
0 |
— |
| 5 |
regime_adaptive_full |
2.00 |
1 |
0 |
— |
| 5 |
regime_adaptive_montage |
2.00 |
1 |
0 |
— |
| 5 |
window_mean_control |
2.00 |
1 |
0 |
— |
| 9 |
ensemble_weighted |
2.33 |
6 |
2 |
afdb_atrial_fibrillation (1), afdb_atrial_fibrillation_multiscale (1) |
| 10 |
coherent_sustained_kuramoto |
2.67 |
3 |
1 |
cap_kuramoto_variants (1) |
| 11 |
synchronisation |
3.00 |
6 |
2 |
afdb_atrial_fibrillation (1), afdb_atrial_fibrillation_multiscale (1) |
| 12 |
critical_slowing_down_baseline |
3.00 |
1 |
0 |
— |
| 13 |
amplitude_gated_delta_kuramoto |
3.33 |
3 |
0 |
— |
| 14 |
transition_entropy |
3.83 |
6 |
0 |
— |
| 15 |
multi_channel_delta_kuramoto |
4.50 |
4 |
0 |
— |
| 16 |
adaptive_channel_kuramoto |
4.67 |
3 |
0 |
— |
| 17 |
snr_weighted_delta_kuramoto |
5.25 |
4 |
0 |
— |
| 18 |
sustained_delta_kuramoto |
6.67 |
3 |
0 |
— |
Cross-domain patterns
Detectors that appear in more than one domain, sorted by mean rank:
critical_slowing_down_multiscale — mean rank 1.00, present in 5 domain(s), wins 5: afdb_atrial_fibrillation_multiscale (1), chb01_seizures_multiscale (1), csd_variant_synthetic (1), dakos_climate_transitions_multiscale (1), psml_grid_oscillation_multiscale (1).normalized_delta_envelope — mean rank 1.25, present in 4 domain(s), wins 3: cap_kuramoto_variants (2), cap_multichannel_staging (1), regime_adaptive_ensemble (1), sleepedf_kuramoto_variants (1).critical_slowing_down — mean rank 2.00, present in 7 domain(s), wins 3: afdb_atrial_fibrillation (3), afdb_atrial_fibrillation_multiscale (4), chb01_seizures (1), chb01_seizures_multiscale (2), dakos_climate_transitions (1), psml_grid_oscillation (1), psml_grid_oscillation_multiscale (2).ensemble_weighted — mean rank 2.33, present in 6 domain(s), wins 2: afdb_atrial_fibrillation (1), afdb_atrial_fibrillation_multiscale (1), chb01_seizures (2), chb01_seizures_multiscale (3), psml_grid_oscillation (3), psml_grid_oscillation_multiscale (4).coherent_sustained_kuramoto — mean rank 2.67, present in 3 domain(s), wins 1: cap_kuramoto_variants (1), regime_adaptive_ensemble (4), sleepedf_kuramoto_variants (3).synchronisation — mean rank 3.00, present in 6 domain(s), wins 2: afdb_atrial_fibrillation (1), afdb_atrial_fibrillation_multiscale (1), chb01_seizures (3), chb01_seizures_multiscale (4), psml_grid_oscillation (4), psml_grid_oscillation_multiscale (5).amplitude_gated_delta_kuramoto — mean rank 3.33, present in 3 domain(s), wins 0: cap_kuramoto_variants (3), regime_adaptive_ensemble (5), sleepedf_kuramoto_variants (2).transition_entropy — mean rank 3.83, present in 6 domain(s), wins 0: afdb_atrial_fibrillation (4), afdb_atrial_fibrillation_multiscale (5), chb01_seizures (4), chb01_seizures_multiscale (5), psml_grid_oscillation (2), psml_grid_oscillation_multiscale (3).multi_channel_delta_kuramoto — mean rank 4.50, present in 4 domain(s), wins 0: cap_kuramoto_variants (5), cap_multichannel_staging (2), regime_adaptive_ensemble (7), sleepedf_kuramoto_variants (4).adaptive_channel_kuramoto — mean rank 4.67, present in 3 domain(s), wins 0: cap_kuramoto_variants (4), regime_adaptive_ensemble (6), sleepedf_kuramoto_variants (4).snr_weighted_delta_kuramoto — mean rank 5.25, present in 4 domain(s), wins 0: cap_kuramoto_variants (6), cap_multichannel_staging (3), regime_adaptive_ensemble (8), sleepedf_kuramoto_variants (4).sustained_delta_kuramoto — mean rank 6.67, present in 3 domain(s), wins 0: cap_kuramoto_variants (7), regime_adaptive_ensemble (9), sleepedf_kuramoto_variants (4).
Ranked refinement backlog
- Advance the
critical_slowing_down_multiscale variant — it wins in 5 early-warning domain(s) (afdb_atrial_fibrillation_multiscale, chb01_seizures_multiscale, csd_variant_synthetic, dakos_climate_transitions_multiscale, psml_grid_oscillation_multiscale) and has the best mean rank (1.00) among detectors present in multiple domains. Extend it to the remaining real-data domains (EEG, cardiac) and compare it head-to-head with the baseline CSD on every corpus.
- Protect and productise
normalized_delta_envelope for CAP sleep staging — it dominates its domain (mean rank 1.25) and should become the default reference detector there.
- Protect and productise
lag1_autocorrelation for synthetic critical-slowing-down corpus — it dominates its domain (mean rank 1.00) and should become the default reference detector there.
- Deprioritise SNR-weighted Kuramoto — in the CAP multichannel panel it does not outperform the unweighted multi-channel Kuramoto detector. Reallocate effort toward channel-selection or coupling-structure variants rather than a raw SNR weighting.
- Audit the
ensemble_weighted fusion rule — it is present in 6 early-warning domains but rarely wins (mean rank 2.33). Investigate whether the current weighting is dominated by a single indicator and whether a learned combination would help.
- Re-evaluate
critical_slowing_down_surrogate — it appears in only one domain and never wins; consider whether the feature is under-powered or simply unsuited to that data regime.
- Re-evaluate
regime_adaptive_full — it appears in only one domain and never wins; consider whether the feature is under-powered or simply unsuited to that data regime.
- Re-evaluate
regime_adaptive_montage — it appears in only one domain and never wins; consider whether the feature is under-powered or simply unsuited to that data regime.
- Re-evaluate
window_mean_control — it appears in only one domain and never wins; consider whether the feature is under-powered or simply unsuited to that data regime.
- Re-evaluate
critical_slowing_down_baseline — it appears in only one domain and never wins; consider whether the feature is under-powered or simply unsuited to that data regime.
Notes
- Detection rate for early-warning aggregates is approximated by
observed_led / n_transitions — the fraction of transitions for which the detector produced a statistically meaningful lead.
- A detector is marked as beating chance when its reported p-value is below 0.05; honest-audit aggregates additionally report the committed
fraction_beats_chance value.
- The CAP multichannel finding that SNR-weighted Kuramoto did not improve over the simple mean-R Kuramoto detector is carried forward explicitly; further investment in that exact spatial-R feature is not supported by the current evidence.