davis-demo.toml

# Demo city configuration — drives `make demo` end to end over the synthetic,
# known-answer fixtures. Pointing nearmiss at a real city is a copy of this file
# with new paths and an exposure layer; no code changes (configurability).
# Paths resolve relative to this file's directory.
city = "Davis"
dataset_note = "Synthetic demonstration data — not real reports. The hotspots are planted in the fixture, so the method finds what was put there."
exposure_unit = "bike trips (synthetic)"

streets = "../tests/fixtures/davis/streets.geojson"
reports = "../tests/fixtures/davis/reports.json"
exposure = "../tests/fixtures/davis/exposure.json"
source_registry = "sources/davis-demo.toml"

raw_dir = "../data/raw/davis"        # PRIVATE, gitignored
out_dir = "../data/published"        # open, committed
submissions_dir = "../data/pending"  # PRIVATE moderation queue, gitignored (HR4)

ref_lat = 38.5449
ref_lon = -121.7405

# Analysis window (METHODOLOGY §1): every published rate is bounded to a stated
# period. This span covers all synthetic fixture timestamps so the demo filters
# nothing and outputs stay byte-stable; a real config would set its true period.
[window]
start = "2026-01-01"
end = "2026-12-31"

[thresholds]
snap_max_m = 25
dedupe_window_s = 600
dedupe_distance_m = 15
small_n = 5
min_publish_n = 3       # k-anonymity: withhold segments with 0 < reports < this
rate_per = 1000
confidence_z = 1.96
fdr_alpha = 0.05        # Benjamini-Hochberg false-discovery-rate level
gi_band_m = 300
kde_bandwidth_m = 150
kde_grid = 20

# EXP-05 prototype (docs/privacy/exp-05-dp-segment-time-bands.md): epsilon-DP segment x
# part-of-day release. Left commented out and disabled: enabling this on real (non-demo)
# data requires a recorded privacy-SME sign-off first (dp_segment_time_release() raises
# without it -- see stats/dp_temporal.py).
# [dp_segment_time]
# enabled = true
# epsilon = 1.0
# sme_signoff_ref = "reviewed by <name>, <date>, see <link to review notes>"