Remake of the Community Severance Index for New Haven, Connecticut. Methodology based on Development of a community severance index for urban areas in the United States: A case study in New York City (Benavides et al, 2024) (journal article, GitHub repository).

• 05/08/2025

  • Apply inverse weighted distance to fill NA values from Connecticut AADT dataset (a_05_regrid_hpms.R).
  • Expand parameter grid to search for best fit PCP with wider range for rank values (b_01_calc_csi.R).
  • Both changes resulted in 2 RM, accoutngin for 100% of variable variance.
  • Upload CSI values (data/github/sf_csi_nh.csv) and plots (figures/ggplot_*).

• 05/07/2025

  • Apply standard scale (mean = 0; sd = 1) to CSI variables.
    • Original scaled only based on sd of each variable and did not account for mean, which caused poor performance in the PCP dimensionality reduction.
    • Update retains 3 RM, accounting for 95% of variable variance
    • See a_09_merge.R Line 76 for updated scale and b_01_calc_csi.R Line 207 for output.

> fa_csi
Factor Analysis using method =  minres
Call: psych::fa(r = list_rrmc_opt$L, nfactors = ., n.obs = num_n, rotate = "varimax",
    scores = "regression")
Standardized loadings (pattern matrix) based upon correlation matrix
                              MR1   MR3   MR2   h2       u2 com
autom_netw_dens              0.14  0.99  0.04 1.00 -0.00297 1.0
autom_inters_dens            0.20  0.95  0.23 0.99  0.01075 1.2
barrier_factor_osm           0.88 -0.10  0.39 0.93  0.06530 1.4
barrier_factor_faf5          0.76  0.43  0.47 0.99  0.00630 2.3
motorway_prox                0.72  0.63 -0.20 0.95  0.04954 2.1
primary_prox                 0.97  0.24  0.02 1.00  0.00057 1.1
secondary_prox               0.60  0.17  0.69 0.87  0.13231 2.1
trunk_prox                   0.77  0.54  0.22 0.94  0.06260 2.0
interstate_highway_prox      0.53  0.34  0.38 0.54  0.45806 2.6
freeways_expressways_prox    0.71  0.67  0.00 0.94  0.05798 2.0
other_princ_arter_prox       0.97  0.22  0.06 1.00  0.00317 1.1
tertiary_prox                0.96  0.19  0.23 1.00 -0.00358 1.2
residential_prox             0.16 -0.09  0.97 0.98  0.02400 1.1
aadt_esri_point              0.53  0.83  0.10 0.97  0.03010 1.7
aadt_fhwa_segm               0.01  0.88  0.37 0.92  0.08381 1.4
co2_emis_per_m2              0.39  0.87  0.29 1.00 -0.00144 1.6
pedest_netw_dens            -0.06  0.31  0.93 0.97  0.03339 1.2
street_no_autom_inters_dens  0.25  0.45  0.86 0.99  0.00589 1.7
NatWalkInd                   0.49  0.64  0.57 0.98  0.02471 2.9

                       MR1  MR3  MR2
SS loadings           7.23 6.43 4.30
Proportion Var        0.38 0.34 0.23
Cumulative Var        0.38 0.72 0.95
Proportion Explained  0.40 0.36 0.24
Cumulative Proportion 0.40 0.76 1.00

Mean item complexity =  1.7
Test of the hypothesis that 3 factors are sufficient.

df null model =  171  with the objective function =  319.67 with Chi Square =  30954.29
df of  the model are 117  and the objective function was  255.02

The root mean square of the residuals (RMSR) is  0.03
The df corrected root mean square of the residuals is  0.04

The harmonic n.obs is  105 with the empirical chi square  34.53  with prob <  1
The total n.obs was  105  with Likelihood Chi Square =  24184.1  with prob <  0

Tucker Lewis Index of factoring reliability =  -0.167
RMSEA index =  1.4  and the 90 % confidence intervals are  1.391 NA
BIC =  23639.58
Fit based upon off diagonal values = 1
Measures of factor score adequacy
                                                   MR1  MR3  MR2
Correlation of (regression) scores with factors   1.00 1.00 0.99
Multiple R square of scores with factors          0.99 0.99 0.99
Minimum correlation of possible factor scores     0.98 0.98 0.98

• Calculate CSI using weighted mean of MR1, MR2, and MR3 scores for each census block group.
  • By updating the scale method of input values, PCP identifies 3 MR which account for 95% of the variance.
  • CSI is calculated by multiplying the weight of each MR (ie. the % variance it explains) by the score, and summing for each census block group.
  • Output object contains this "raw" CSI value ($csi), as well as normalized ($csi_normal) , and normalized values * 100 for interpretability ($cs_100)

> df_scores[, c("csi", "csi_normal", "csi_100")]
# A tibble: 105 × 3
        csi csi_normal csi_100
      <dbl>      <dbl>   <dbl>
 1  0.00259     0.159    15.9
 2  0.0656      0.176    17.6
 3 -0.159       0.118    11.8
 4  0.192       0.208    20.8
 5 -0.379       0.0618    6.18
 6 -0.126       0.127    12.7
 7 -0.351       0.0689    6.89
 8 -0.269       0.0900    9.00
 9 -0.311       0.0792    7.92
10  0.364       0.252    25.2
# ℹ 95 more rows
# ℹ Use `print(n = ...)` to see more rows

• 05/01/2025
  • Change spatial context from New Haven neighborhoods shapefile with sf::st_union to town boundary shapefile sourced from Connecticut DEEP ArcGIS archive (https://deepmaps.ct.gov/datasets/CTDEEP::connecticut-and-vicinity-town-boundary-set/about?layer=1).
  • When plotted, the unioned neighborhoods (red) have a different border than the town boundary (black), which would result in a different spatial context and different values for interpolation to census block groups and calculation of environmental variables.
  • Direct query to OpenStreetMaps with post filtering instead of SQL filter at query time.

a_00_initiate.R

• Install and load required packages.
• Set home and input/output directories and Connecticut-specific coordinate reference system.
• Source at the beginning of each run script source("scripts/a_00_initate.R")

a_01_prep_sld.R

• Prepare spatial data from Smart Location Database.
• Input(s): data/input/SmartLocationDatabase.gbd
• Output(s): data/output/a_01/sf_sld_variables_proj.rds, data/output/a_01/df_sld.rds

a_02_prep_barrier.R

• Prepare barrier factor variables.
• Input(s): data/input/connecticut-latest.osm.pbf, data/input/FAF5Network/F5FNEWHAVEN.shp
• Output(s): data/output/a_02/sf_barrier_factors_nh.rds

a_03_query_traffic.R

• Query US (AADT) and Connecticut (HPMS) traffic data from ESRI respoitories.
• Output(s): data/output/a_03/sf_aadt_proj.rds, data/output/a_03/sf_ct_hpms_proj.rds

a_04_regrid_aadt.R

• Regrid (interpolate) traffic intensity (counts) data to census block groups with local regrid_ok().
• Input(s): data/output/a_01/sf_sld_variables_proj.rds, data/output/a_03/sf_aadt_proj.rds, R/regrid_ok.R
• Output(s): data/output/a_04/sf_grid_nh.rds, data/output/a_04/sf_regrid_aadt_cbg.rds

a_05_regrid_hpms.R

• Regrid (interpolate) traffic intensity (segements) data to census block groups with local regrid_ok()
• Input(s): data/output/a_04/sf_grid_nh.rds, data/output/a_03/sf_ct_hpms_proj.rds, R/regrid_ok.R
• Output(s): data/output/a_05/sf_regrid_hpms_cbg.rds

a_06_prep_osm.R

• Prepare OpenStreetMaps data.
• Output(s): data/input/geofabrik_connecticut-latest.osm.pbf, data/input/geofabrik_connecticut-latest.gpkg, data/output/a_06/sf_osm_nh_filter.rds

a_07_calc_road.R

• Calculate road infrastructure statistics for census block groups.
• Input(s): data/output/a_04/sf_grid_nh.R, data/input/FAF5Network/F5F_NEWHAVEN.shp, data/output/a_06/sf_osm_nh_filter.rds
• Output(s): data/output/a_07/sf_grid_prox.rds

a_08_calc_emissions.R

• Calculate traffic CO2 emissions per m^2 for census block groups.
• Input(s): data/input/DARTE_v2.gdb, data/output/a_04/sf_grid_nh.rds
• Output(s): data/output/a_08/sf_co2_emis.rds

a_09_merge.R

• Merge all census block group-level variables into a single object.
• Input(s): All .rds files generated in scripts a_00 to a_08.
• Output(s): data/output/a_09/sf_csi_scale.rds, data/output/a_09/df_csi_geoid_scale.rds

b_01_calc_csi.R

• Calculate Community Severance Index (CSI) for New Haven census block groups.
• Input(s): data/output/a_01/df_sld.rds, data/output/a_09/sf_csi_scale.rds
• Output(s): data/output/b_01/sf_csi_nh.rds, data/output/b_01/sf_csi_nh.csv, data/github/sf_csi_nh.csv

c_01_plot_csi.R

• Plot CSI values for publication figures.
• Input(s): data/output/b_01/sf_csi_nh.rds
• Output(s): data/output/c_01/sf_csi_polygons.rds, figures/ggplot_csi_faf5.png, figures/ggplot_barrier_faf5.png

d_01_calc_tree.R

• Calculate tree cover statistics for census block groups.
• Input(s): data/input/greenness/tree_cover_New_Haven.tif
• Output(s): data/output/d_01/sf_tree.rds

d_02_calc_no2.R

• Calculate NO2 statistics for census block groups.
• Input(s): data/input/no2/HAQ_TROPOMI_NO2_CONUS_QA75_L3_Annual_2019_V2.4_20240729.nc4
• Output(s): data/output/d_02/sf_no2.rds

d_03_calc_pm25.R

• Calculate PM2.5 statistics for census block groups.
• Input(s): data/input/pm2.5/V5GL0502.HybridPM25.NorthAmerica.201901-201912.nc
• Output(s): data/d_03/sf_pm25_proj.rds

d_04_calc_temperature.R

• Calculate Cooling Degree Day statistics for census block groups.
• Input(s): data/daymet/tmax/daymet_v4_daily_na_tmax_2019.nc, data/daymet/vp/daymet_v4_daily_na_vp_2019.nc
• Output(s): data/output/d_04/sf_cdd.rds

d_05_query_demographic.R

• Query demographic data from United States census with tidycensus package.
• Output(s): data/output/d_05/sf_demographic_nh.rds

d_06_merge.R

• Merge census block group-level CSI values with environmental conditions data.
• Input(s): data/output/d_01/...rds
• Output(s): data/output/d_06/sf_csi_cond.rds

e_01_simple.R

• Analyze relationships between CSI and environmental/demographic conditions with simple linear regression methods.
• Input(s): data/output/d_06/sf_csi_cond.rds

e_02_simple_yj.R

• Analyze relationships between Yeo-Johnson transformed CSI and environmental conditions with simple linear regression methods.
• Input(s): data/output/d_06/sf_csi_cond.rds

e_03_mgvc.R

• Explore mgvc models for CSI and environmental/demographic conditions.
• Input(s): data/output/d_06/sf_csi_cond.rds

• Deprecated R scripts which are not needed for analysis.
• Sensitivity analyses are not required for our analysis as the sensitivity analyses conducted in the original Benavides et al, 2024 paper were conducted to support the 19 included variables.
• Our analysis replicates the supported methodology and included variables and therefore does not need its own sensitivty supports.