Loading approved evidence
Retrieving and verifying the requested analytical release. Result data loads before optional map boundaries and source checks.
Your filters and release stay in the address. Navigation remains available.
Retrieving and verifying the requested analytical release. Result data loads before optional map boundaries and source checks.
Your filters and release stay in the address. Navigation remains available.
STATS19 · Clock-change study · Phases 1–3
Compare the weeks around the autumn and spring clock changes. Keep collision frequency, injury severity and young-driver involvement separate.
CURRENT CONCLUSION
Great Britain; 16 development years: 2005–2019 and 2022. Before/after collision counts, severity proportions and driver involvements are separate outcomes. No per-mile denominator is supplied for these comparisons. 2023–2025 are reserved for this study, but were used elsewhere.
Read the bands correctly. No decile ranking. Autumn and spring are separate comparisons. The primary window is 21 days either side, excluding transition Sunday; some national diagnostic charts use their own explicitly stated scope.
1 / Question
Separate autumn and spring and distinguish collision counts from fatal proportions and young-driver involvement. Each answers a different question.
Read the frozen question →2 / Measure
Audit linked records and zero-collision hours. Exclude transition Sunday to avoid mixing 23-hour and 25-hour days with ordinary days.
Inspect the data audit →3 / Initial pattern
Use matched whole-week windows and inspect individual years. A pooled pattern may hide inconsistent annual behaviour.
Explore descriptive patterns →4 / Adjustment
The current release cannot separate the transition from weather, season, holidays and traffic. The protocol defines the later counterfactual test.
Read the planned adjustment →5 / Validation
Between-year intervals describe uncertainty in the raw comparison. They do not validate an adjusted causal effect or an exposure-normalised risk change.
Inspect remaining dependencies →6 / Decision
The supported conclusion is descriptive. Obtain the missing inputs before estimating the adjusted effect or making a pricing interpretation.
Return to the recorded conclusion →Loading the frozen development evidence…