Skip to contents

Compute Weighted Percentiles cross a List of Data Frames (with optional grouping)

run_weighted_percentiles.Rd

Applies the compute_weighted_percentiles() function to each data frame in a named list, using a specified variable and optional weight. Supports optional grouping via a categorical by variable, and returns either percentile values or share values, depending on the share argument.

Usage

run_weighted_percentiles(
  data_list,
  var_name,
  wgt_name = NULL,
  probs = seq(0, 1, 0.25),
  type = c("type_4", "type_2"),
  share = FALSE,
  na.rm = TRUE,
  by = NULL
)

Arguments

data_list

A named list of data frames, (e.g., across countries or years).

var_name

A string specifying the variable name (e.g., "dhi", "pilabour") to compute percentiles or shares for.

wgt_name

An optional string specifying the weight variable to be used. If NULL, equal weights are assumed.

probs

A numeric vector of probabilities with values between 0 and 1, defining percentiles (if share = FALSE) or the brackets between which shares are computed (if share = TRUE).

type

A character string indicating which percentile definition to use. #' Either "type_4" (default, linear interpolation-based of the empirical cdf - continuous sample quantile) or "type_2" (used in Stata commands like collapse and _pctile, inverse of empirical distribution function with averaging at discontinuities - discontinuous sample quantile).

share

Logical. If TRUE, returns income shares between percentile brackets instead of the percentile values. Default is FALSE. Note: This always uses definition = "type_4" (interpolation), regardless of the definition parameter. It cannot be combined with definition = "type_2".

na.rm

Logical. If TRUE, missing values in var_name or wgt_name are removed.

by

Optional string giving the name of a categorical variable to split the data within each data frame before computing statistics.

Value

A named list of data frames. If by is used, unless if share = FALSE and length(probs) ==1, the result will be a nested list where the outer list is by data frame and the inner list is by subgroup. If by is NULL, the result is a simple list of numeric vectors (one per data frame).

Details

Percentiles are computed using weighted version of quantile definition 4 from Hyndman and Fan (1996), by default, or quantile definition 2 if specified. When share = TRUE, the function estimates Lorenz ordinates by taking quantiles from the running sum of the ordered outcome variable (divided by the total), according to the same quantile definition 4 only.

Examples

if (FALSE) { # \dontrun{  

# Import data
my_data_list <- lissyuse(data = "es", from  = 2016)

# Retrieve the percentile estimates
percentiles_result <- my_data_list %>% 
  run_weighted_percentiles(
    var_name = "dhi",
    wgt_name = "hpopwgt",
    probs = seq(0.1, 0.9, 0.1),
    type = "type_2",
    na.rm = TRUE
)
print(percentiles_result)

# Compute the distribution shares
shares_result <- run_weighted_percentiles(
    data_list = my_data_list,
    var_name = "dhi",
    wgt_name = "hpopwgt",
    probs = seq(0, 1, 0.1),
    share = TRUE,
    na.rm = TRUE
)
print(shares_result)

# Using the by option 
by_result_median <- run_weighted_percentiles(
    data_list = purrr::map(my_data_list[1:2], ~.x %>% filter(emp == 1)),
    var_name = "pi11",
    wgt_name = "ppopwgt",
    probs = 0.5,
    type = "type_4",
    na.rm = TRUE, 
    by = "region_c"
   
)
print(by_result_median)
} # }