logtagrosscon ~ nbhd + X + Y + trend + UAF + elev +
(1 + STY | BLOCK) + (1 + logsf + floor_extract| ADDRESS)Condo Sales Model
Mixed Model
Using sales from 2022-2024 in Manhattan, Brooklyn, and Queens, and sales from 2020-2024 in the Bronx and Staten Island to predict the sale price of condos across NYC, the most successful mixed model (or hierarchical linear model) to date is the following:
Citywide Model Deciles
These tables show the metrics divided by decile on a model run on the entire city population simultaenously, rather than boro-by-boro. Below are deciles overall, deciles by boro, and deciles by nbhd.
Citywide Overall Deciles
Citywide Model Deciles by Boro
Manhattan
The Bronx
Brooklyn
Queens
Staten Island
assessr::
Citywide Model Deciles By NBHD
The following shows the metrics broken down by decile for each nbhd in the city.
Boro-Specific
The interactive map below visualizes key assessment accuracy metrics at the neighborhood level across New York City. Toggle between the different layers which shade neighborhoods by different performance indicators, such as: R, Spearman’s Rank, Coefficient of Dispersion (COD), Price-related Differential (PRD), and Modified Kakwani Index (MKI). Clicking on a neighborhood reveals a popup with detailed statistics for that area.
This map displays every neighborhood with at least two qualified sales in the boro-specific time period. While the same model was applied to every borough, it was done so individually, and not as a citywide population.
create_metrics_map(
data = neighborhood_shapefile,
by = "FullNbhdName",
center_lng = -73.99,
center_lat = 40.76,
zoom_level = 11
)
This map only displays neighborhoods with at least 10 qualified sales in the boro-specific period.
create_metrics_map(
data = min10,
by = "FullNbhdName",
center_lng = -73.99,
center_lat = 40.76,
zoom_level = 11
)