Week 3#
Spatial and temporal patterns of vehicle strikes#
A Red Dot Is Not a Research Finding#
Last week, this week#
- week 2: reporting practices created an apparent peak in time
- this week: reporting practices can create an apparent hotspot in space
- same failure mode, different axis
The initial map#
Schematic aggregate view — bar length only, not a real road network.
A red dot is not a research finding. Twelve red dots are not automatically twelve times more convincing.
The raw ranking#
| Road | Raw collisions |
|---|---|
| Coorabin Road | 18 |
| Grellan Creek Rd / Merricourt Hwy | 14 |
| Wombaroo Pass | 13 |
| Old Bindari Rd / Tindarra Loop Rd | 11 |
| Bunjalung Ridge Road | 9 |
Congratulations: you have found the road with the most records. Whether it is the most dangerous road remains under review.
The denominator problem#
- raw count needs no denominator — which is exactly the problem
- per kilometre: divide by road length
- per million vehicle-kilometres: divide by traffic as well
The map has excellent visual authority and no opinion about your denominator.
Worked exposure calculation#
collisions / (traffic × days × length km) × 1,000,000
| Quantity | Coorabin Rd (busy) | Old Bindari Rd (quiet) |
|---|---|---|
| Collisions | 18 | 11 |
| Traffic (ADT) | 2,800 | 220 |
| Length | 38 km | 9 km |
| Days observed | 1,409 | 1,409 |
| Rate / million veh-km | 0.12 | 3.94 |
Fewer collisions, thirty times the rate. Same formula, same period, opposite conclusion.
The revised ranking#
Same seven roads, ranked by rate instead of count. Old Bindari Road moves from the middle of the pack to the top.
Reporting coverage#
| Coverage | Roads | What limits reporting |
|---|---|---|
| High | Coorabin Rd, Merricourt Hwy | Patrols, cameras |
| Medium | Bunjalung, Grellan Ck, Tindarra | Irregular patrol, carer network |
| Low | Old Bindari Rd, Wombaroo Pass | No routine patrol, patchy phone coverage |
Old Bindari Road has both the highest rate and the lowest coverage. Exposure adjustment does not eliminate reporting bias — it corrects the denominator, not who was watching.
Choosing a hotspot definition#
A road is a hotspot if its rate is at least double the median rate across the roads being compared.
- median rate: 1.06 — threshold: 2.13
- one road clears it: Old Bindari Road
- a different threshold would draw the line somewhere else — that’s why you state it
The peer challenge#
Swap your ranking and definition with another pair. Ask of theirs:
- does a road get excluded that traffic and reporting coverage suggest it shouldn’t be?
- is a road included mainly because it’s well reported, not because it’s dangerous?
Revised conclusion — and next week#
The raw ranking and the exposure-adjusted ranking name different roads as the top concern. Both are defensible measures of different things; only one of them accounts for how much a road is actually used.
A road with fewer collisions can still be the more concerning road. Next week asks a different question again: not which road, but what a collision actually does once it happens — the physics of speed and impact.