Skip to contents

Aggregate block-group ACS indicators up to tract, county, and state

Source: R/calc_acs_by_geography.R
calc_acs_by_geography.Rd

Part (c) of issue #395: roll block-group ACS demographic values up to tract, county, and state for the EJSCREEN web app's "additional demographics" (and side-by-side) layers, using the same per-indicator aggregation rules as doaggregate().

Usage

calc_acs_by_geography(
  bg = blockgroupstats,
  id_col = "bgfips",
  levels = c("blockgroup", "tract", "county", "state"),
  pop_fallback = TRUE,
  exclude_islandareas = TRUE,
  out_dir = NULL
)

Arguments

bg

Block-group table with id_col and ACS indicator columns. Defaults to blockgroupstats. Not modified (a copy is made internally).

id_col

Block-group FIPS column name. Default "bgfips"; also used as the id column name in the block-group-level output. If the column is numeric, its leading zero is restored before parent GEOIDs are derived from it.

levels

Any of "blockgroup", "tract", "county", "state".

pop_fallback

If TRUE (default), weighted-mean indicators whose weight column is missing from the data fall back to population weighting, as in doaggregate(). If FALSE, such indicators are dropped.

exclude_islandareas

If TRUE (default), drop Island Area block groups (AS/GU/MP/VI/UM) before aggregating – identified by the ST column via is.island() when present (their FIPS are non-standard length). Set FALSE to keep them (they will have 0/NA demographics).

out_dir

Optional directory; if set, each level is written there as acs_by_<level>.csv.

Value

A named list of data.frames, one per requested level: the geography FIPS id column, the summed count columns, and the recomputed weighted-mean indicator columns.

Details

Each indicator is aggregated by the method EJAM stores for it, looked up with the same accessors doaggregate() uses:

  • calctype() == "sum of counts" -> the column is summed within each geography (population, households, the ACS universes/denominators, raw counts, etc.).

  • calctype() == "wtdmean" -> the column is a weighted mean, where the weight is the indicator's own denominator from calcweight() (for example pop for population fractions, hhlds / occupiedunits / builtunits for household-based indicators, povknownratio for the low-income fraction, age25up for educational attainment, and so on). The mean is sum(value * weight) / sum(weight), taken only over block groups that have the indicator.

Following doaggregate(), if an indicator's weight column is not present in the data, the population weight pop is used as a fallback (with a message). Percentile (lookedup), ratio to avg, and map bin/text columns are not recomputed here – those need geography-level distributions / lookups and are a separate step (see issue #395). The geography id is the leading FIPS digits of bgfips: 12 = block group, 11 = tract, 5 = county, 2 = state.

Island Areas (American Samoa, Guam, the Northern Mariana Islands, the U.S. Virgin Islands) have no standard ACS demographics, so by default (exclude_islandareas = TRUE) their block groups are dropped before aggregating. This matches EJAM's v3 decision that Island Area demographics are unavailable by default (rather than emitting them with na.rm sums of 0), and matches EPA's ACS layer, which omits them. Puerto Rico is not an Island Area for this purpose (it has standard ACS demographics) and is retained.

See also

doaggregate() calctype() calcweight() is.island() calc_ejscreen_export() calc_ejscreen_threshold_layers()