Synthetic U.S. Spatiotemporal Covid-19 Datasets Generated Through USPopulationSampler Package

Dataset Summaries

This repository serves as the official data companion to the USPopulationSampler package. It provides standardized, pre-processed datasets designed to facilitate high-resolution geographic sampling, demographic analysis, machine learning, and epidemiological modeling. When the package is loaded into R, users have access to the following datasets described below: 

1. Synthetic Spatiotemporal Covid-19 Datasets 

Description: A collection of 28 simulated epidemiological datasets designed to mimic the statistical properties, temporal trends, and transmission dynamics of the actual pandemic that were generated by the USPopulationSampler package stored as parquet files. Each dataset comprises of randomly generated and independent spatiotemporal data with the locations being generated uniformly in census blockgroups which themselves are randomly selected according to population counts within a given county.  Datasets were generated from 5/27/2026 to 5/28/2026. 

The dimension of each synthetic data is 100,501,034 rows and 5 columns with each row representing a case of Covid-19. The data columns include longitude, latitude, date of each case from 2020 to 2023, the fips identifier code, and the geo_id. 

Number of variables: 5 

Number of cases/rows: 100,501,034

Variable List: 

2. Processed Covid-19 Dataset 

Description: A curated, cleaned historical record of the Covid-19 pandemic in the United States, spanning from its onset in 2020 through 2023. The original raw data was taken from the https://github.com/nytimes/covid-19-data. 

Number of variables: 4

Number of cases/rows: 100,501,034

Variable List:

3. USPopulationSampler Reference Data 

Description: Embedded directly within or complemented by the package, this dataset provides a highly granular geographic and demographic baseline of the United States based on the official published 2020 Decennial Census - Census Block Maps. The data goes down to the block group level which is the smallest geographic unit for which the Census Bureau publishes population data).

Number of variables: 9

Number of cases/rows: 242,335

Specialized formats or other abbreviations used:

FIPS = Federal Information Processing Standards geographic code identifying U.S states and territories and counties.

GEOID = Census Geographic Identifier. Values are Census Block Group identifiers in the format 1500000USSSCCCTTTTTTB, where:

• ss = state FIPS code (2-digit)
• ccc = county FIPS code (3-digit)
• TTTTTT = census tract code (6-digit)
• B = blockgroup number

lat = Latitude in decimal degrees (WGS84)

lon = Longitude in decimal degrees (WGS84)