data trove food desert states summary

Overview

Source: /home/coolhand/html/datavis/data_trove/data/quirky/food_desert_states.json

Saturn profiled 51 rows across 11 columns. The stats below are deterministic and machine-readable; the prose is a language-model interpretation of those stats (opt-in, added after the fact, never sees raw rows).

name categorical label

This column contains US state names, with all 51 entries being unique (cardinality = 51, n = 51), consistent with a full list of US states plus Washington D.C. or a territory. Entropy ratio is exactly 1.0, meaning perfect uniformity — every value appears exactly once (top_rate = 0.0196, or 1/51). The 'long_tail' alert is technically correct but misleading here: the distribution is not skewed, it is perfectly flat.

Treatment: Use as a categorical label or join key for state-level lookups; one-hot encode or ordinal-map if used as a model feature.

abbr categorical identifier

This column contains two-letter US state abbreviations, with exactly 51 unique values across 51 rows — covering all 50 states plus one additional entry (likely Washington D.C. or a territory). Every value appears exactly once (top_rate = 0.0196), yielding a perfect entropy ratio of 1.0, meaning this is a fully uniform identifier with zero redundancy. The 'long_tail' alert is a statistical artifact of perfect uniformity, not a genuine concern here.

Treatment: Use as a join key or primary identifier for state-level lookups; one-hot encode or map to region groupings if used as a feature.

pop numeric feature

This column likely represents population counts for 51 distinct geographic or administrative units (e.g., U.S. states or territories), given exactly 51 fully unique, non-null integer values. The distribution is heavily right-skewed (skew = 2.58, kurtosis = 7.61), with a median of 4,372 far below the mean of 6,338 and a maximum of 38,643 — suggesting a small number of very large-population entities pulling the tail; 4 outliers (~7.8% of rows) drive this effect. The std of 7,243 exceeds the mean, confirming high dispersion relative to the central tendency.

Treatment: Log-transform before regression or distance-based modelling to reduce skew and outlier influence.

desertPop numeric feature

This column likely represents a population count associated with desert regions (e.g., population living in desert areas, possibly by U.S. state or similar unit given n=51). The distribution is severely right-skewed (skew=4.73, kurtosis=25.51): the median is just 21.0 while the mean is 38.27 and the max reaches 449.0, indicating a small number of entities dominate desert population totals. With 6 outliers (≈11.8% of rows) and a standard deviation of 67.39 against a median of 21.0, those extreme values will heavily distort any linear model trained on raw values.

Treatment: Log-transform (or apply sqrt) before modelling to reduce skew; investigate and cap or flag the 6 outliers separately.

povertyPop numeric feature

This column likely represents a count of people living in poverty, measured per U.S. state (n=51, matching the 50 states plus DC). The distribution is heavily right-skewed (skew=2.53, kurtosis=6.80), with a median of 548 but a mean of 794 and a maximum of 4685, indicating a small number of high-population states pull the mean well above the typical value. Four outliers (~7.8% of rows) are flagged, likely corresponding to the most populous states with the largest absolute poverty populations. The near-uniqueness (49 of 51 distinct values) suggests this is a genuine count variable, not a derived category.

Treatment: Log-transform before regression to reduce skew and mitigate outlier influence.

noVehicle numeric feature

This column likely represents a count of households or individuals without access to a vehicle, aggregated at some geographic unit (e.g., census tract or neighbourhood) across 51 observations. The distribution is severely right-skewed (skew = 4.41, kurtosis = 22.57), with a median of 115 but a mean pulled to 204.9 by a long upper tail reaching 2202. Six outliers (≈11.8% of rows) are driving this extreme shape, suggesting a small number of densely populated or car-deprived areas dominate the upper end while most units cluster between 40 and 203 (IQR = 163).

Treatment: Log-transform or apply a robust scaler before modelling; investigate the 6 outlier units for data quality or genuine extreme values.

povertyRate numeric feature

This column represents poverty rate (likely percentage of population below the poverty line) across 51 observations — almost certainly U.S. states plus DC. Values range from 7.33 to 19.2 with a mean of 12.35 and median of 11.91, indicating a modest right skew (skew=0.75) consistent with a handful of higher-poverty states pulling the tail. Three outliers (~5.9% of rows) at the upper end are flagged, likely representing the highest-poverty states; the near-zero kurtosis (0.20) suggests the distribution is otherwise fairly normal.

Treatment: Use as-is or apply mild log-transform to reduce right skew before regression; investigate 3 upper outliers for leverage effects.

noVehiclePct numeric feature

This column represents the percentage of households without a vehicle, likely a census or survey-derived socioeconomic indicator across 51 geographic units (e.g., states or counties). The distribution is heavily right-skewed (skew=4.53, kurtosis=21.72) with the bulk of values tightly clustered between Q1=2.15% and Q3=3.05%, yet 3 outliers pull the max to 17.37% — more than 5× the median of 2.45%. That extreme upper tail almost certainly reflects a high-density urban area (e.g., New York City) where car-free households are far more common than in typical units.

Treatment: Cap or Winsorize at the 95th percentile before modelling, or log-transform to compress the extreme upper tail.

counties numeric feature

This column most likely represents the number of counties per U.S. state (plus D.C.), matching the dataset's 51 rows exactly. The mean of ~62 and median of 62 are consistent with typical state county counts, while the maximum of 254 is almost certainly Texas (which has 254 counties). The distribution is right-skewed (skew 1.44) with high kurtosis (3.91), driven by that single outlier — Texas — which sits far above the rest of the distribution.

Treatment: Use as-is for regression/analysis; consider log-transform to reduce skew caused by the Texas outlier.

lat numeric feature

This column contains latitude coordinates, almost certainly representing the 50 US states plus Washington D.C. (n=51, all unique). The mean of 39.57 and median of 39.55 are tightly aligned, indicating near-symmetric distribution centered on the mid-continental US, though a kurtosis of 3.94 flags heavier tails than normal — driven by the 2 outliers likely corresponding to Alaska (max 64.2) and Hawaii (min 19.9).

Treatment: Use as-is or pair with a longitude column for geospatial modelling; consider flagging Alaska and Hawaii as geographic outliers if contiguous-US analysis is intended.

lon numeric feature

This column contains longitude coordinates, almost certainly representing geographic locations of 51 entities (e.g., US states or cities), all with negative values indicating the Western Hemisphere. The range spans -155.58 to -69.45, consistent with continental US plus Hawaii (≈-155°), and the left skew (skew = -1.27) reflects Hawaii and Alaska pulling the distribution westward. Three duplicate longitude values exist (51 records, 48 unique), and 2 outliers (~3.9%) likely correspond to Hawaii and/or Alaska.

Treatment: Use as-is for spatial analysis or pair with latitude for geographic modelling; consider flagging the 2 outlier values (Hawaii/Alaska) if contiguous-US-only analysis is needed.