The explorer now includes three Quiesce.T screening tools in addition to the raw indicators. Each tool is designed for transparent public planning: raw values are cleaned, converted to comparable rates where appropriate, normalized to 0–100 percentiles, combined into weighted domains, and assigned to five priority tiers only when enough data are present.
01Data sources
Inputs include ACS 2020–2024 housing and tenure estimates, NYC housing-maintenance and DOB enforcement records, eviction activity, property-market indicators, affordable-housing program records, MapPLUTO parcel attributes, and NYC Flood Vulnerability Index climate scenarios.
Why essential: no single dataset captures instability, preservation need, and climate vulnerability. The tools combine complementary public signals while keeping source measures available for inspection.
02Cleaning and validation
Blank strings, nulls, NA-like text, infinities, and missing API values are treated as missing. They are not converted to zero. Source-specific identifiers are standardized before joins, and tract values are preserved even when a climate or administrative layer has no match.
Why essential: zero means a measured absence; missing means the source did not report a usable value. Confusing the two can falsely create high or low priority scores.
03Rates before rankings
Counts are converted to denominated rates when a defensible denominator exists, such as evictions per 1,000 renter households, open violations per 1,000 renter households, and market transactions per 1,000 renter households. Percentage measures are used directly.
Why essential: large areas naturally have more events. Rate construction allows tracts and districts of different sizes to be compared more fairly.
04Outlier treatment
Highly skewed rates are transformed with log1p and winsorized at the 1st and 99th percentiles before percentile ranking. This limits the influence of a few extreme observations without deleting them from the map.
Why essential: administrative housing data often contain extreme counts. A transparent cap prevents one outlier from setting the entire citywide color scale.
05Normalization
Every valid input is transformed into a 0–100 citywide percentile where 100 always means greater need, risk, or priority. Reverse-direction indicators are inverted only when explicitly defined.
Why essential: violations, rent burden, FVI, and vacancy are measured in different units. Percentiles put them on a common scale while preserving relative position.
06Weighted domains
Housing Priority combines affordability, displacement pressure, housing distress, and tenant vulnerability. Housing Preservation combines physical distress, aging stock, market pressure, affordability/preservation, and tenant vulnerability. Climate & Resident Safety combines flood vulnerability, unsafe housing, and resident sensitivity.
Why essential: domains keep the tool interpretable. A score can be traced back to the kind of problem driving it rather than becoming a black-box average.
07Missing-data gates
A score is published only when at least 75% of weighted inputs are available and every required domain has usable data. Climate & Resident Safety also requires at least one Flood Vulnerability Index input. Otherwise the area is labeled Insufficient Data.
Why essential: missing data should not produce false precision. Areas with weak coverage are visible but not ranked.
08Composite score and tiers
Domain scores are combined by documented weights, then percentile-ranked citywide. Published scores run from 0 to 100 and are grouped into Tier 1 Very High Priority, Tier 2 High Priority, Tier 3 Elevated Priority, Tier 4 Moderate Priority, and Tier 5 Lower Priority.
Why essential: tiering makes the map easier to read while preserving the continuous score in the profile and table.
09Geographic processing
Census tracts are the base analytical geography. Community Districts and state legislative districts use tract-derived summaries. Parcels show current MapPLUTO attributes plus the tract/district context rather than pretending that every tract-level indicator is parcel-specific.
Why essential: geographic aggregation choices affect results. The site identifies when values are native and when they are tract-derived summaries.