trim_weights: Trim weights

In pewresearch/pewmethods: Pew Research Center Methods Miscellaneous Functions

Description

Trim survey weights. Wrapper for the trimWeights() function in the survey package that eliminates the need to create a svydesign object.

Usage

trim_weights(
  weight,
  lower_quantile = 0.01,
  upper_quantile = 0.99,
  minval = NULL,
  maxval = NULL,
  strict = FALSE
)

Arguments

weight	A numeric vector of weights
lower_quantile	The quantile at which the weights should be trimmed on the lower end
upper_quantile	The quantile at which the weights should be trimmed on the upper end
minval	NULL by default. If the lower quantile is less than the minimum value, minval will be used for the trim.
maxval	NULL by default. If the upper quantile is greater than the maximum value, maxval will be used for the trim.
strict	When the trimmed weights are reapportioned, some values may exceed the specificed threshold value. If strict = FALSE this is permitted. Otherwise, weights are trimmed recursively until none exceed the limit.

Value

A numeric vector of trimmed weights

Examples

calculate_deff(dec13_excerpt$weight)
trimmed <- trim_weights(dec13_excerpt$weight, lower_quantile = 0.1, upper_quantile = 0.9)
# Trimming reduces the standard deviation of the weights, reducing reducing the design effect
# and increasing the effective sample size.
calculate_deff(trimmed)

pewresearch/pewmethods documentation built on March 27, 2020, 7:22 p.m.