saturn

This dataset covers 3,222 U.S. counties with housing-unit counts (owner-occupied, renter-occupied, total) plus a FIPS code, county name, and the percent of renters. The three count columns are extremely right-skewed (skew between 9.5 and 15.8, kurtosis above 140) with 13–14% of rows flagged as outliers — a handful of huge urban counties (max total_housing_units of about 3.36M vs a median of roughly 10,021) dominate the distribution. The pct_renter field is far better behaved, centered near 26% with a much tighter spread, making it the most useful comparable metric across counties. Start by inspecting the long tail of total_housing_units, then use pct_renter to compare counties on a normalized basis.

This column is the FIPS code for U.S. counties — every one of 3,222 rows is unique with no nulls, matching the count of U.S. counties. Values span 1001 to 72153, consistent with state-prefixed county FIPS identifiers, and the distribution is essentially uniform across the code space (skew 0.157, kurtosis -0.63, no outliers).

This column holds fully-qualified US county names (e.g. 'X County, State'), with 3222 rows all unique and zero nulls. The token 'county,' appears 2999 times, so roughly 223 rows use a different administrative suffix (parish, borough, census area). Texas (256), Virginia (189), and Georgia (159) lead the state distribution, consistent with the real US county count.

Counts of total housing units per record, almost certainly at a county or similar geographic level given 3,222 rows with 3,074 unique values and no nulls. The distribution is severely right-skewed (skew 12.05, kurtosis 240.5) with a median of 10,021 but a max of 3,363,093, and 443 rows (13.7%) flagged as outliers well above the Q3 of 25,939. The mean of 39,402 sits far above the median, confirming a long heavy tail driven by a few very large geographies.

This appears to be a count of owner-occupied housing units per geographic area, with 3001 unique values across 3222 rows and effectively no zeros (zero_rate 0.0003) or nulls. The distribution is severely right-skewed (skew 9.52, kurtosis 146.9): the median is 7325.5 but the mean is 25551.7 and the max reaches 1,552,164, producing 429 outliers (13.3% outlier rate). The interquartile range (3147.75 to 18863.5) is dwarfed by the standard deviation of 67553, indicating a long tail of large jurisdictions.

Counts of renter-occupied housing units per record, ranging from 28 to 1,810,929 with a median of 2,579.5 — consistent with a geographic rollup (likely county or similar). The distribution is extremely right-skewed (skew 15.82, kurtosis 398.15) and 13.9% of rows fall outside the IQR fences, reflecting a few very large metros dominating a long tail of small areas. No nulls or zeros, and 2,709 unique values across 3,222 rows.

This is a numeric feature representing the percentage of renters per record, ranging from 3.01 to 100.0 with a mean of 27.35 and median of 26.07. The distribution is right-skewed (skew 1.32, kurtosis 4.41) with 88 outliers (2.7%) on the high end, suggesting a small set of records — likely dense urban areas — with renter shares far above the typical 21.64–31.66 IQR. No nulls or zeros, and 1925 unique values across 3222 rows indicate well-populated continuous data.