Cassandra Lab

Self-discovering leading-indicator engine — mining cross-vertical event chains for validated "A precedes B" rules.

Tested 9 hypotheses. Treat lift on small n with caution; we report Wilson 95% lower bound + base rate so significance is self-evident.

NIH grant precedes clinical trial start

nih_grant → trial_start · window 540d

need more history

n=0hits=0 lift=—

n=0 < min sample 8

Trial completion precedes SEC 8-K (public sponsor disclosure)

trial_complete → sec_8k · window 120d

weak

n=17hits=0 P=0.0% base=2.5% lift=0× conf(95% low)=0.0%

Federal award precedes contractor SEC 8-K (material event)

fed_award → sec_8k · window 90d

need more history

n=0hits=0 lift=—

n=0 < min sample 8

Federal (research) award precedes trial start

fed_award → trial_start · window 365d

need more history

n=0hits=0 lift=—

n=0 < min sample 8

Complaint velocity precedes CPSC recall

complaint → recall · window 120d

need more history

n=0hits=0 lift=—

n=0 < min sample 8

FDA recall precedes SEC 8-K

fda_recall → sec_8k · window 90d

weak

n=134hits=0 P=0.0% base=1.9% lift=0× conf(95% low)=0.0%

CPSC recall precedes SEC 8-K

recall → sec_8k · window 90d

weak

n=179hits=0 P=0.0% base=1.9% lift=0× conf(95% low)=0.0%

NIH grant precedes sponsor SEC 8-K

nih_grant → sec_8k · window 540d

need more history

n=0hits=0 lift=—

n=0 < min sample 8

Trial start precedes trial completion (sanity baseline)

trial_start → trial_complete · window 1095d

need more history

n=0hits=0 lift=—

n=0 < min sample 8