We are developing a software package based on our Virginia project (see projects page). The software starts with American Community Survey Public Use Microdata (PUMS) data for a Public Use Microdata Area (PUMA) and produces synthetic data for census tracts and block groups. The synthetic data is just like the PUMS data. The output file has individual records in rows and variables in columns. Census tract and block group detail tables (B, S and P tables) are used to adjust the marginals of a PUMS PUMA table/model. The output data has weights (optionally replicate weights) and the total population (sum of weights) is equal to the population of the census tract or blockgroup. The “regular” ACS PUMS data can be merged with the Supplemental Poverty Measure PUMS data so variables from that dataset can be used, for example whether or not the individual receives WIC benefits. The tract level data sets can be “stacked/merged” to create synthetic data for cities, towns, counties, school districts, and other geographies that might not be the same as a PUMA geography.
Several models are built in and specify the variables in the PUMA model as well as the detail tables used for the adjustment that produces the tract/block group synthetic data. The user can define their own models and adjusting marginals by writing a function that specifies the variables used in the PUMA model as well as the marginal tables used for adjustment.
The package uses the “R” statistical computing language. The computation uses Iterative Proportional Fitting (IPF) to produce the synthetic data. The population in the census tract/block group matches the total population for that geography. A C computer language function/subroutine does that fitting and produces about a 20% increase in speed and uses about 1/2 the computer memory when compared to the Haberman AS 51 1972 Fortran subroutine. The basic R function has the same arguments as the Ipfp function in the mipfp R package.
NEWS: 31 March 2024: I have posted an updated 0.9 version Windows installer. The updated version handles replicate weights that take into account the Margin of Error for the adjusting (marginal) ACS tables. The documentation needs to be updated. Email if you want to use this feature (see Contact Us page)
I have posted a Windows installer for the package on GitHub. See link below. See the README. I have also posted 3 vignettes as pdfs on the main GitHub page so you can read them without downloading and installing the package. (12-12-2023).
I am improving the user interface and adding documentation pages. Email info@dorerfoundation.org to get on the email list for updates. (11-23-2023)
Link to GitHub website
Supplemental Poverty Measure Data (.RData format)
2021
2019
2018
2017
American Community Survey (ACS) Data Page
Census Supplemental Poverty Measure
R mipfp package
mipfp package journal article
Photo Credit: John Adams birthplace in Quincy Massachusetts. On the family kitchen table in this house, John Adams drafted the Commonwealth of Massachusetts constitution. It is the oldest written constitution in continuous use. Photo by Ray Stafford copyright 2023 used with permission.