Skip to contents

Calibrate survey weights to known population totals

Source: R/calibrate.R
calibrate.Rd

Adjusts survey weights so that the weighted marginal totals match known population values. Supports linear (GREG) and logit calibration methods for categorical auxiliary variables.

Usage

calibrate(
  data,
  variables,
  population,
  weights = NULL,
  wt_name = "wts",
  method = c("linear", "logit"),
  type = c("prop", "count"),
  control = list(maxit = 50, epsilon = 1e-07)
)

Arguments

data

A data.frame, weighted_df, survey_taylor, or survey_nonprob. survey_replicate → error. Any other class → error.

variables

<tidy-select> Columns to calibrate on. Must be categorical (character or factor). Specify as a bare name or c(var1, var2, ...).

population

Named list of population targets. Names must match the column names selected by variables. Each element: a named numeric vector c(level = target, ...).

For type = "prop": values must sum to 1.0 (within 1e-6 tolerance). For type = "count": values must be strictly positive.

weights

<tidy-select> Weight column name (bare name). NULL → auto-detected from weighted_df attribute or survey object @variables$weights. For plain data.frame with weights = NULL, uniform starting weights are used and the output column is named by wt_name (default "wts").

wt_name

Character scalar. Name of the output weight column in the returned weighted_df. Default "wts". Ignored when data is a survey object (survey_taylor or survey_nonprob).

method

Character scalar. "linear" (default): one-step exact GREG calibration (may produce negative weights). "logit": bounded iterative calibration (always positive).

type

Character scalar. "prop" (default): population values are proportions. "count": population values are counts.

control

Named list of convergence parameters. Merged with defaults list(maxit = 50, epsilon = 1e-7) — omitted keys retain their defaults.

Value

  • data.frame or weighted_df input → weighted_df

  • survey_taylor or survey_nonprob input → same class as input (survey_taylor or survey_nonprob; class is preserved)

The weight column in the output contains calibrated weights. A history entry with operation = "calibration" is appended to weighting_history.

See also

Other calibration: poststratify(), rake()

Examples

df <- data.frame(
  age_group = c("18-34", "35-54", "55+", "18-34", "35-54"),
  sex = c("M", "F", "M", "F", "M"),
  stringsAsFactors = FALSE
)
pop <- list(
  age_group = c("18-34" = 0.30, "35-54" = 0.40, "55+" = 0.30),
  sex = c("M" = 0.48, "F" = 0.52)
)
result <- calibrate(df, variables = c(age_group, sex), population = pop)