Where the data comes from
Current results come from the Pennsylvania Department of Environmental Protection, Michigan Department of Environment, Great Lakes, and Energy, Minnesota Department of Health, Wisconsin Department of Health Services, Colorado Department of Public Health and Environment, and Washington State Department of Health, and from the CDC Environmental Public Health Tracking Network for Iowa. Pennsylvania county figures are calculated from nonnegative individual test-result records; Wisconsin and Colorado provide county aggregates. Washington publishes county test counts by radon level for each year from 2000; we add the years together and calculate shares from the summed counts, and leave out its yearly means and medians, which are rounded and cannot be combined. Michigan publishes results by ZIP code for first-time tests in homes without an active mitigation system; we sum ZIP codes with full statistics into the county holding most of each ZIP code's housing units, using the Census ZCTA-county relationship file, and do not combine medians across ZIP codes. Iowa figures come from pre-mitigation tests that laboratories report to the CDC Tracking Network; shares are calculated from CDC's county counts by radon level, and medians are CDC's published county values. Minnesota publishes county percentages, exact threshold counts, and average annual properties tested, but not an exact total denominator. We preserve those published measures and do not estimate the missing total. Counties in 40 other states use the same CDC laboratory data as Iowa; counties where CDC suppresses any count are shown without figures. States with no verified dataset (Hawaii, Kansas, Mississippi and New Jersey) are not presented as having results.
Source refreshes
Source updates are loaded through state-specific import scripts. Each load validates county FIPS codes, numeric values, record counts, and source attribution before replacing the corresponding live records. A failed load leaves the previously published records in place.
Sample sizes & confidence
- 30+ tests: the county statistic can be shown as a headline figure.
- Fewer than 30 tests: the county page does not present the result as a standalone headline statistic.
- Missing values remain unavailable rather than being estimated from a neighboring county, state average, or predicted EPA zone.
One source per number
A figure never blends sources. Where a county has several, the primary figure is the source with the latest period end (larger test count breaks ties). The others appear in an "Other data for this county" table.
Selection-bias caveat
These are voluntary and transaction tests, not random samples. That's stated on every data page. Units are pCi/L throughout, with Bq/m³ alongside (1 pCi/L = 37 Bq/m³).
Attribution on every figure
County pages display the source name, reporting period, sample size, and a link to this methodology. Source metadata in the database also records the publisher and source URL used by the import process.