Research Methodology

Purpose

HOV Injury Intelligence is a public-interest research initiative operated by HOV Law, PLLC. Its purpose is to make Central Florida crash and traffic-safety data easier to understand, explore, and cite — for residents, journalists, researchers, and public agencies. It is editorially separate from the firm's legal-marketing content: research pages state facts with sources, or state plainly that data is not yet available. The center never publishes estimated, extrapolated, or placeholder statistics.

Data collection

Data comes exclusively from the published sources listed on the data sources page, each with its integration status, access date, coverage, and known limitations. Sources are ingested one at a time, never silently combined; where two sources have different jurisdictional coverage, any combined analysis will document exactly how they were merged.

Data cleaning

The following normalization steps are implemented and applied to every ingested record:

• Road-name normalization — whitespace cleanup and a maintained alias table so corridor variants (e.g., “SR 50,” “E Colonial Dr”) count as one road. The alias table is versioned with this methodology.
• Date and time normalization — the source-reported crash date and local clock time are validated and preserved with the America/New_York time-zone label; unparseable or future dates are rejected and counted, not guessed.
• Flag normalization — involvement flags (pedestrian, bicycle, motorcycle, commercial vehicle) map Y/N/1/0 variants to true/false; an absent flag becomes unknown, never false.
• Coordinate screening — coordinates outside a Florida bounding box are treated as geocoding errors: the record is kept for counts, but its coordinates are removed so maps never plot it. Nulled coordinates are counted and reported.
• Duplicate handling — records with a previously seen source identifier are removed and counted as duplicates in the validation summary.

Every ingestion produces a published validation summary (records imported, valid, rejected by reason, duplicates removed, coordinates nulled). No record is silently discarded.

Definitions

Crash

A reported traffic crash record in the connected dataset. Reported crashes undercount actual incidents: minor collisions frequently go unreported.

Fatal crash

A crash in which the source recorded at least one fatality. Sources following Florida/KABCO conventions code this as severity “K.”

Serious injury crash

A crash whose most severe recorded outcome is an incapacitating injury (KABCO “A”). Definitions of “incapacitating” follow the source agency's coding manual, not our judgment.

Pedestrian / bicycle / motorcycle / commercial vehicle crash

A crash where the source recorded the corresponding involvement flag as true. When a source does not carry a flag, we report “no data,” never zero.

Intersection crash

A crash record carrying both a primary road and a cross street after normalization. Records without both fields cannot be attributed to an intersection and are excluded from intersection rankings (they remain in totals).

Ranking methodologies

Rankings use explicit, named methodologies; every ranked page states which one it uses. Implemented methodologies:

• By total crashes — total reported crashes per road or intersection; ties broken by fatal crashes, then serious-injury crashes, then name.
• By fatal crashes — fatal crashes as the primary metric.
• By serious-injury crashes — serious-injury (incapacitating) crashes as the primary metric.
• By crash rate per centerline mile — total crashes divided by verified roadway length. This methodology refuses to run when verified mileage is missing, rather than producing a partial ranking. It is not yet used on any published page.

Count-based rankings do not adjust for traffic volume or road length; a longer, busier corridor will record more crashes. There is deliberately no composite “danger score.” If one is ever introduced, every factor and weight will be published here and versioned.

Limitations

• Reported crashes are not a complete record of all incidents; unreported collisions are invisible to this analysis.
• Recent periods are subject to reporting delays; source agencies revise records after initial publication.
• Jurisdictional coverage differs by source (state-maintained roads vs. all public roads; city vs. county reporting).
• Geocoding is imperfect; some crashes cannot be located precisely, and location-based analyses state how many records they exclude.
• Severity coding follows the source agency's manual and its known inter-agency inconsistencies.
• Crash data cannot support legal conclusions about any individual collision.

Updates

Datasets are refreshed when source agencies publish updates, and each analysis page shows the period it covers. Methodology changes are versioned; material changes will be noted here with the date and the reason.

Sources referenced by this methodology

1. [1] Florida Crash Dashboard & Traffic Crash Reports, Florida Department of Highway Safety and Motor Vehicles (FLHSMV). https://www.flhsmv.gov/traffic-crash-reports/ Integration planned — not yet ingested.
2. [2] Signal Four Analytics, University of Florida GeoPlan Center. https://signal4analytics.com/ Integration planned — not yet ingested.
3. [3] FDOT Open Data Hub, Florida Department of Transportation (FDOT). https://gis-fdot.opendata.arcgis.com/ Integration planned — not yet ingested.
4. [4] Fatality Analysis Reporting System (FARS), National Highway Traffic Safety Administration (NHTSA). https://www.nhtsa.gov/research-data/fatality-analysis-reporting-system-fars Accessed 2026-07-11.