Skip to contents

Configure replicate weights for variance estimation

Source: R/utils.R
add_replicate.Rd

This function configures replicate weights (bootstrap, jackknife, etc.) that allow estimation of variance for complex statistics in surveys with complex sampling designs. It is essential for obtaining correct standard errors in population estimates.

Usage

add_replicate(
  weight,
  replicate_pattern,
  replicate_path = NULL,
  replicate_id = NULL,
  replicate_type
)

Arguments

weight

String with the name of the main weight variable in the survey (e.g., "pesoano", "pesomes")

replicate_pattern

String with regex pattern to identify replicate weight columns. Examples: "wr\d+" for columns wr1, wr2, etc.

replicate_path

Path to the file containing replicate weights. If NULL, assumes they are in the same main dataset

replicate_id

Named vector specifying how to join between the main dataset and replicate file. Format: c("main_var" = "replicate_var")

replicate_type

Type of replication used. Options: "bootstrap", "jackknife", "BRR" (Balanced Repeated Replication)

Value

List with replicate configuration that will be used by the sampling design for variance estimation

Details

Replicate weights are essential for:

  • Correctly estimating variance in complex designs

  • Calculating appropriate confidence intervals

  • Obtaining reliable coefficients of variation

  • Performing valid statistical tests

The regex pattern must exactly match the replicate weight column names in the file. For example, if columns are named "wr001", "wr002", etc., use the pattern "wr\\d+".

This function is typically used within add_weight() for more complex weight configurations.

See also

add_weight for complete weight configuration svrepdesign for replicate design in survey package load_survey where this configuration is used

Examples

if (FALSE) { # \dontrun{
# Basic configuration with external file
annual_replicates <- add_replicate(
  weight = "pesoano",
  replicate_pattern = "wr\\\\d+",
  replicate_path = "bootstrap_weights_2023.xlsx",
  replicate_id = c("ID_HOGAR" = "ID"),
  replicate_type = "bootstrap"
)

# With replicates in same dataset
integrated_replicates <- add_replicate(
  weight = "main_weight", 
  replicate_pattern = "rep_\\\\d{3}",
  replicate_type = "jackknife"
)

# Use within add_weight
weight_config <- add_weight(
  annual = add_replicate(
    weight = "pesoano",
    replicate_pattern = "wr\\\\d+", 
    replicate_path = "bootstrap_annual.xlsx",
    replicate_id = c("numero" = "ID_HOGAR"),
    replicate_type = "bootstrap"
  ),
  monthly = "pesomes"  # No replicates for monthly
)

# In load_survey
ech <- load_survey(
  path = "ech_2023.dta",
  svy_type = "ech",
  svy_edition = "2023",
  svy_weight = add_weight(
    annual = add_replicate(
      weight = "pesoano",
      replicate_pattern = "wr\\\\d+",
      replicate_path = "bootstrap_weights.xlsx", 
      replicate_id = c("numero" = "ID"),
      replicate_type = "bootstrap"
    )
  )
)
} # }