Skip to contents

Post-stratify survey weights to known joint population cell totals

Source: R/poststratify.R
poststratify.Rd

Adjusts survey weights so that the weighted cell counts (or proportions) match known population values for every joint combination of stratification variables. Unlike calibrate() and rake(), which match marginal totals, poststratify() matches exact cross-tabulation cells in a single pass.

Usage

poststratify(
  data,
  strata,
  population,
  weights = NULL,
  wt_name = "wts",
  type = c("prop", "count")
)

Arguments

data

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

strata

<tidy-select> Stratification variables that jointly define the cells. Specify as a bare name or c(var1, var2, ...). Unlike calibrate() and rake(), strata variables may be any type (character, factor, integer, numeric).

population

A data.frame with one column per variable selected by strata (column names must match exactly), one column named "target", and one row per unique cell combination.

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

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.

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).

type

Character scalar. "prop" (default): target values are proportions summing to 1.0. "count": target values are population counts. Consistent with calibrate() and rake().

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 post-stratified weights. A history entry with operation = "poststratify" is appended to weighting_history.

See also

Other calibration: calibrate(), rake()

Examples

df <- data.frame(
  age_group = c("18-34", "35-54", "55+", "18-34", "35-54", "55+"),
  sex = c("M", "M", "M", "F", "F", "F"),
  stringsAsFactors = FALSE
)

# Proportion targets (default type = "prop")
pop_prop <- data.frame(
  age_group = c("18-34", "35-54", "55+", "18-34", "35-54", "55+"),
  sex = c("M", "M", "M", "F", "F", "F"),
  target = c(0.14, 0.18, 0.17, 0.15, 0.19, 0.17)
)
result <- poststratify(df, strata = c(age_group, sex), population = pop_prop)

# Count targets (explicit type = "count")
pop_count <- data.frame(
  age_group = c("18-34", "35-54", "55+", "18-34", "35-54", "55+"),
  sex = c("M", "M", "M", "F", "F", "F"),
  target = c(14000, 18000, 17000, 15000, 19000, 17000)
)
result2 <- poststratify(df, strata = c(age_group, sex),
  population = pop_count, type = "count")