iterake: Iterative raking procedure
In ttrodrigz/iterake: Tools for iterative raking

Description Usage Arguments Value Examples

View source: R/iterake.R

This function creates row-level weights using an iterative raking algorithm based on targets from a known population (established with universe()). The weights are appended as a new column in the data. If iterake() converges, the weighted marginal proportions of the sample will match those set in universe(). Summary statistics of the weighting procedure are presented by default.

1
2
3

iterake(universe, wgt.name = "weight", max.wgt = 3,
  threshold = 1e-10, max.iter = 50, stuck.limit = 5,
  permute = FALSE, summary = TRUE)

`universe`	Output object created with `universe()` function.
`wgt.name`	Name given to column of weights to be added to data, default is "weight", optional.
`max.wgt`	Maximum value weights can take on, default is 3, optional. The capping takes place prior to applying expansion factor (if `N` is set in `universe()`.
`threshold`	Value specifying minimum summed difference between weighted marginal proportions of sample and universe before algorithm quits, default is 1e-10, optional.
`max.iter`	Value capping number of iterations for the procedure, default is 50, optional.
`stuck.limit`	Value capping the number of times summed differences between sample and universe can oscillate between increasing and decreasing, default is 5, optional.
`permute`	Boolean indicating whether to test all possible orders of categories in `universe` and keep the most efficient (`TRUE`) or to test categories in the order listed in `universe` only (default, `FALSE`), optional. Note that when `TRUE` this will increase runtime by a factor of `(number of categories)!`.
`summary`	Whether or not to display summary output of the procedure, default is `TRUE`, optional.

Data frame with the resulting weight variable appended to it.

data(demo_data)

iterake(
    universe = universe(
        data = demo_data,
        
        category(
            name = "Sex",
            buckets = factor(
                x = levels(demo_data[["Sex"]]),
                levels = levels(demo_data[["Sex"]])
            ),
            targets = c(0.4, 0.5),
            sum.1 = TRUE
        ),

        category(
            name = "BirthYear",
            buckets = c(1986:1990),
            targets = rep(0.2, times = 5)
        ),
    
        category(
            name = "EyeColor",
            buckets = c("brown", "green", "blue"),
            targets = c(0.8, 0.1, 0.1)
        ),
    
        category(
            name = "HomeOwner",
            buckets = c(TRUE, FALSE),
            targets = c(3/4, 1/4)
        )
    )
)