eiCompare (version 3.0.0)

precinct_agg_combine: Aggregates racial estimates across geographic units

Description

Obtains aggregated precinct counts of racial groups from a voter file. This function is usually applied after application of BISG, when the voter file has probabilistic estimates of race. However, it can be applied more generally, aggregating actual counts of race. This function can perform aggregation over probabilistic estimates of race and ground truth race at the same time.

Usage

precinct_agg_combine(
  voter_file,
  group_col = "precinct",
  race_cols = NULL,
  true_race_col = NULL,
  true_race_keys = NULL,
  include_total = FALSE
)

Arguments

voter_file

The voter file, as a dataframe. Should contain columns that denote the race probabilities or the actual race of the voter.

group_col

A string denoting the column to aggregate over (e.g., "precinct").

race_cols

A list of strings denoting which columns contain probabilistic estimates of race. By default, it assumes output from WRU. This function does not require all WRU output columns be present in the voter file; rather, it checks which outputs are present and uses those in aggregation.

true_race_col

A string denoting which (single) column in the voter file specifies the true race of the voter. If this variable is provided, then true_race_keys must also be provided, or an error is thrown.

true_race_keys

A named list, with keys denoting the new race groups (e.g., "white", "black", "hispanic", etc.). The value of each key is a string or list of strings that denote which columns in the voter file map onto the new ground truth race column. This is useful, for example, in mapping multiracial and Native American voters onto an "other" race category. This variable should only be provided if aggregating over the true race.

include_total

A logical denoting whether the total counts (potentially rounded) should be included in the output dataframe.

Value

Aggregated dataset of nrow() precinct size, including racial size precinct estimates. Dataset suitable for EI/RxC.

Examples

# NOT RUN {
# Create synthetic voter file with typical BISG output
voter_file <- data.frame(
  precinct = c(1, 1, 2, 2),
  pred.whi = c(0.10, 0.20, 0.30, 0.40),
  pred.bla = c(0.40, 0.30, 0.20, 0.10),
  pred.his = c(0.10, 0.20, 0.30, 0.40),
  pred.asi = c(0.30, 0.20, 0.10, 0.00),
  pred.oth = c(0.10, 0.10, 0.10, 0.10)
)
# Function uses these column names by default
agg <- precinct_agg_combine(
  voter_file = voter_file,
  group_col = "precinct",
  include_total = FALSE
)

# Running aggregation with a ground truth race column
voter_file <- data.frame(
  precinct = c(1, 1, 1, 1, 2, 2, 2, 2),
  race = c("BL", "WH", "NA", "MR", "BL", "WH", "BL", "BL")
)
# Need to specify race keys for true race column
agg <- precinct_agg_combine(
  voter_file = voter_file,
  group_col = "precinct",
  true_race_col = "race",
  true_race_keys = list("whi" = "WH", "bla" = "BL", "oth" = c("NA", "MR")),
  include_total = TRUE
)

# Running aggregation for both predicted and true race columns. Note the
# change in column names, which means we need to specify column names.
voter_file <- data.frame(
  precinct = c(1, 1, 2, 2),
  p.whi = c(0.10, 0.20, 0.30, 0.40),
  p.bla = c(0.40, 0.30, 0.20, 0.10),
  p.his = c(0.10, 0.20, 0.30, 0.40),
  p.asi = c(0.30, 0.20, 0.10, 0.00),
  p.oth = c(0.10, 0.10, 0.10, 0.10),
  race = c("BL", "WH", "BL", "WH")
)
agg <- precinct_agg_combine(
  voter_file = voter_file,
  group_col = "precinct",
  race_cols = c("p.whi", "p.bla", "p.his", "p.asi", "p.oth"),
  true_race_col = "race",
  true_race_keys = list("whi" = "WH", "bla" = "BL"),
  include_total = FALSE
)
# }