saturn

This dataset contains 3,222 US county-level records with four fields: county name, FIPS code, Gini index, and state. The Gini index is the most analytically interesting column, with a mean of 0.448 and a max of 0.721, plus 56 outliers worth investigating for unusually high local inequality. The state distribution is broad (52 unique values), led by Texas (254 counties) and Georgia (159), so any state-level comparison should account for that imbalance. County names show a 39% duplicate rate, reflecting common names like Washington, Jefferson, and Franklin County that recur across states.

This is the U.S. county FIPS code: a 5-digit numeric identifier where the first two digits encode state and the last three encode county. With 3222 unique values across 3222 rows, no nulls, and a range from 1001 to 72153 spanning the standard FIPS state prefixes, every row corresponds to a distinct county. Distribution stats (mean 31377, std 16299, near-zero skew) are artifacts of the prefix encoding and not meaningful as a numeric feature.

This column holds US county-level place names: nearly every value ends in 'County' (2999 of 3222 rows), with smaller contingents of 'Parish' (64, Louisiana), 'Municipio' (78, Puerto Rico), and 'City' (47). Heavy duplication is expected and present — 39.2% duplicate rate with 1262 repeats — because common names like Washington, Jefferson, and Franklin County recur across states. Lengths are tight (10–46 chars, mean 14.2, ~2 words) and there are no nulls or empties.

This is a US state code field with 52 distinct values across 3222 rows and no nulls, consistent with the 50 states plus DC and likely a territory. Distribution closely tracks county counts: TX leads at 254 (7.88%), followed by GA (159) and VA (133), and entropy is high at 5.31 (ratio 0.93), indicating broad spread rather than concentration. The 52-value cardinality is the only mild surprise—worth confirming whether the extras are DC, PR, or stray codes.

Numeric column holding Gini index values, all within the plausible 0.2744–0.721 range with no nulls or zeros across 3222 rows. The distribution is tight (IQR 0.049, std 0.038) and centred near 0.448, but a mild right skew (0.50) and 56 high-end outliers (1.7%) suggest a handful of unusually unequal observations.