Method: Proportional, resident-estimate-adjusted parcel unit counts — redistribute_parcel_pums

A wrapper around the redistribute function, that takes PUMS data as additional input, and uses it to calculates an adjusted weight for redistribution to parcel data.

Usage

redistribute_parcel_pums_adj(source, target, households,
  target_total = "Total_Units", target_indicator = "Unit_Type",
  households_size = "size", households_indicator = "BLD",
  households_id = "SERIALNO", person = NULL,
  person_household_id = "SERIALNO", ...)

Arguments

source, target: Source and target of redistribute.
households: PUMS household data. Can be a list with household and person entries, like that returned from download_census_pums.
target_total: Column name in target (e.g., parcel data; or equivalent vector) containing the unit count
target_indicator: Column name of a logical variable in target (or equivalent vector) to use to assign adjustment values, where TRUE indicates a single family home. If the values of the column are characters, anything not matching "MULTI" will be TRUE.
households_size: A vector of household sizes, or a column in households containing such values. If this is not specified or found, person will be used to calculate household size, based on hosuehold_id and person_household_id.
households_indicator: Same a target_indicator, but in households, or a vector (or column name) of BLD codes, where either "02" or "03" would be TRUE.
households_id: A vector of household IDs aligning with households, or a column name in households containing such IDs. Only used to calculate households_size if necessary.
person: PUMS person data. Only used to calculate households_size if necessary.
person_household_id: A vector of household IDs aligning with person, or a column name in person containing such IDs. Only used to calculate households_size if necessary.
...: Additional arguments to pass to redistribute. You will likely need to specify a map, and potentially source_id and/or target_id.

Value

Result of redistribute. These are assumed to be parcel-level estimates, which could then be aggregated to make higher-level estimates.

Details

It is assumed that initial weights are unit counts, and these are to be adjusted based on PUMS household data.

Examples

if (FALSE) {
if (require("tidycensus")) {
  # download source, target, and household data
  tracts <- tidycensus::get_acs(
    year = 2021,
    state = "51",
    county = "013",
    geography = "tract",
    output = "wide",
    variables = c(total = "B01001_001"),
    geometry = TRUE
  )
  parcels <- sf::st_read(paste0(
    "https://arlgis.arlingtonva.us/arcgis/rest/",
    "services/Open_Data/od_MHUD_Polygons/",
    "FeatureServer/0/query?where=1=1&outFields=*&outSR=4326&f=json"
  ))
  pums <- download_census_pums(tempdir(), "51", geoids = tracts$GEOID)

  # calculate map between source and target
  map_tr_to_parcel <- redistribute(
    tracts, parcels,
    target_id = "OBJECTID", return_map = TRUE
  )

  # redistribute tract-level summaries to parcels,
  # using a resident-estimate-adjusted weight
  parcels_filled <- redistribute_parcel_pums_adj(
    tracts[, -2], parcels, pums,
    map = map_tr_to_parcel, target_id = "OBJECTID"
  )

  # this can also be calculated with the underlying function

  ## calculate resident estimates
  household_size <- tapply(
    table(pums$person$SERIALNO)[pums$household$SERIALNO],
    pums$household$BLD %in% c("02", "03"),
    mean,
    na.rm = TRUE
  )
  residents <- parcels$Total_Units * household_size[
    as.character(parcels$Unit_Type != "MULTI")
  ]

  ## supply as weights
  parcels_filled_manual <- redistribute(
    tracts[, -2], parcels, map_tr_to_parcel,
    target_id = "OBJECTID", weight = residents
  )

  # either way, you could now use the result to make estimates
  # at higher-level geographies by redistributing from the
  # parcel-level estimates to the new target
}
}