by Colton Zier, Pavel Chernyavskiy, and the University of Wyoming Advanced Research Computing Center.
Case counts were obtained from United States county-level data compiled by the Johns Hopkins Center for Systems Science and Engineering. All output is provided as-is with no explicit or implicit warranty.
This map shows the number of new cases per 100,000 persons reported over the past 3 days.
Care is required for incidence rates in sparsely-populated areas: small populations increase uncertainty and may produce unrealistically-large rates.
Trend evaluation of sparsely-populated counties can be misleading - please interpret with caution.
If the 7-day trend in cases/day is > 20% lower than the 30-day trend in cases/day, we call the trend 'Improved'.
If the 7-day trend in cases/day is > 20% higher than the 30-day trend in cases/day, we call the trend 'Deteriorated'.
If the 7-day trend in cases/day is within 20% of the 30-day trend in cases/day, we call the trend 'Within 20%'.
This map is useful for understanding regional patterns of COVID-19 incidence.
For example, we can use this map to track geographic differences of incidence rates for a specific state, across a collection of states, or across the continental United States. Sparsely-populated counties tend to have uncertain incidence rates. To mitigate this issue, for each county, our statistical model borrows information from its neighbors to improve prediction and reduce variability in rates due to chance alone.
Our model uses several measured characteristics at the county level (number of persons per household, 9-level metric for rurality, median, and elevation of the county centroid) along with state effects, and spatially-correlated county effects to produce predicted rates.
These rates are less uncertain and tend to be more stable than using the reported number of cases alone. We re-estimate our models every day using cases reported over the previous 7 days.