In donboyd5/microweight: Tools for Weighting Microdata Files

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

options(tibble.print_min = 4L, tibble.print_max = 4L)

library(microweight)
library(tidyverse)
library(knitr)

"Reweighting" is the process of adjusting the weights on a microdata file so that the file, with new weights, better represents the population of interest.

The function reweight creates these new weights.

Setting up our data

In this vignette we'll use small microdata files extracted from the American Community Survey in the United States: acs (10,000 individuals) and acsbig (100,000 individuals).

To keep things simple, we create a minimal subset with only these columns:

Column | Description ------------- | ------------- pwgtp | person weight agep | person's age mar | marital status (1=married, 2=single, ...) pincp | personal income, total wagp | wage income ssp | Social Security income

The first few records have a lot of zeros so we'll look at the tail, which is more representative of typical records:

acs2 <- acs %>%
  select(pwgtp, agep, mar, pincp, wagp, ssp)
tail(acs2)

Look at weighted totals for a few variables:

totals <- acs2 %>%
  summarise(across(c(pincp, wagp, ssp), ~sum(. * pwgtp)))
totals

A simple reweighting problem

Suppose that we have other information, perhaps for a later year, and that we want the following targets:

Column | Current value | Targeted value ------------- | -----------------|---------------- pincp | 7,990,814,687 | 8,000,000,000 wagp | 6,070,651,037 | 6,020,123,456 ssp | 546,119,219 | 550,100,000

We want new weights that will hit these targets. To do that we must define the following:

Item | Description ------------- | ------------- iweights | vector of initial weights, here pwgtp target_names | vector of names of the target variables targets | vector of values for the targets tol | vector of tolerances around the target values xmat | matrix with values

Set up the problem

Let's define the first four, as they are easy to understand:

iweights <- acs2$pwgtp
target_names <- c("pincp", "wagp", "ssp")
targets <- c(8e9, 6020123456, 550.1e6)
tol <- c(100, 100, 100)

xmat is simply a matrix of values:

xmat <- acs2 %>%
  select(all_of(target_names)) %>%
  as.matrix()

Solve the problem

Now we are ready to reweight:

result <- reweight(iweights, targets, target_names, tol, xmat)

Examine results

How did it work and how long did it take (in seconds)?

result$solver_message
result$etime

How do the targeted values compare to the original values? reweight returns a list with target_df that includes the following information:

Item | Description --------------- | ------------- targnum | target number targname | target name target | targeted value targinit | value for the target using initial weights targcalc | value calculated with new weights targinit_diff | difference between target and initial value targtol | tolerance allowed around the target targcalc_diff | difference between calculated value with new weights and target targinit_pdiff | percent difference between target and initial value targtol_pdiff | percent difference allowed, based upon tolerances targcalc_pdiff | percent difference between calculated value and target

result$targets_df %>%
  kable(digits=c(rep(0, 8), 2, 2, 2),
        format.args=list(big.mark = ","))

Note that the actual difference between the target for wagp (wages) was $104 even though we set a tolerance of $100. That is because the optimizer that chooses new weights (in this case) has its own set of tolerances. Those can be changed by passing options to the solver, but we won't do that in this vignette.

donboyd5/microweight documentation built on Aug. 17, 2020, 4:48 p.m.