Fireballs
Negative Control1,052 events. If our method says fireballs trigger each other, it's broken. It doesn't.
Decision Chain: Fireballs
Ingest NASA fireball data. 1,052 events with second-level timestamps.
1,052 events ingested. Second-level precision.
Fit self-exciting point process. Fireballs (random meteor events) should show zero self-excitation.
Full-dataset branching ratio 0.960. This is WRONG — fireballs don't trigger each other.
Investigate why the full-dataset ratio is so high. Look for instrument or detection changes.
1994 satellite upgrade dramatically improved detection capability. Pre-1994 vs post-1994 shows a clear structural break.
Refit the Hawkes process within the post-1994 era only, where detection capability is stable.
Era-corrected branching ratio: 0.007. Near-zero self-excitation, as expected.
Negative control: PASSED — and proves era-splitting is mandatory
Fireballs are random meteor entry events. They cannot trigger each other. Without era correction, the branching ratio is 0.960 — suggesting almost total self-excitation. This is entirely artificial, caused by a 1994 satellite upgrade that increased detection capability. After era correction: 0.007. Near-zero, exactly as physics predicts. This result proves two things: (1) our method correctly identifies independent events when they exist, and (2) era-split analysis is mandatory for any dataset with known instrument changes.