NYC street treatment explorer

Every treatment drawn with its published geometry. Three errors in early versions of this pipeline, each since corrected, came from treating a geometry as something it is not: invisible in a table, obvious on this map.

loading

Analysis
Area whole city
type a borough or a street name
Treatments
Combine at one corner
within 20 m
Why one corner can carry two records

A turn calming record and a record on a published daylighting list can be one physical change filed twice, so counting both overstates how much of the city is treated. A ring marks a corner whose records come from more than one source.

Every individual treatment, coloured by what was installed. The published lists are under Collections.
Collections
A collection is a set of locations rather than a treatment type. Loading one restricts the map to its records.
Data sources
Where the records come from, with an editorial trust tier from the provenance graph. Setting a source aside removes its records from the map and from the live estimate.
Context: curb already clear
Hydrants and bus stops prohibit parking by statute, so this curb was clear before any programme reached it. Context rather than treatments: no date, no estimate.
Subclass filters
Placement
Junction rule
Installed on or before
2026

Drag to pan, scroll to zoom, hover for a record.

Estimate for this selection

How this is computed
θ^=1n∑i=1n(Δi−1|Ci|∑j∈CiΔj)
\hat\theta \;=\; \frac{1}{n}\sum_{i=1}^{n}\left(\Delta_i \;-\; \frac{1}{|C_i|}\sum_{j\in C_i}\Delta_j\right)
The published estimator: the mean over treated sites of the change at the site minus the mean change at its own comparison locations.
LaTeX source\hat\theta \;=\; \frac{1}{n}\sum_{i=1}^{n}\left(\Delta_i \;-\; \frac{1}{|C_i|}\sum_{j\in C_i}\Delta_j\right)
Placebo details
Descriptive before and after ratio
RRagg=∑iTafteri/∑iTbeforei∑iCafteri/∑iCbeforei
\text{RR}_{\text{agg}} \;=\; \frac{\sum_i T^{\text{after}}_i \big/ \sum_i T^{\text{before}}_i}{\sum_i C^{\text{after}}_i \big/ \sum_i C^{\text{before}}_i}
The descriptive aggregate rate ratio a browser can compute for any selection, carrying no unit effect and no calendar month effect.
LaTeX source\text{RR}_{\text{agg}} \;=\; \frac{\sum_i T^{\text{after}}_i \big/ \sum_i T^{\text{before}}_i}{\sum_i C^{\text{after}}_i \big/ \sum_i C^{\text{before}}_i}

The ratio can disagree in sign with the modelled estimate: for hardened daylighting at a junction the aggregate ratio is about 0.72 and the modelled estimate is about 1.15.

Installations the window admits

Crashes in the selection

Selection breakdown
Modelled estimates the manuscript reports

Ordered by p ascending, excluded cells last.

The tags, the flags and the bar
Why the controls on the left do not move these rows