Calibrate: 'Calibrate'

In Displayr/flipData: Functions for extracting and describing data

Calibrate

Description

Creates a sampling weight based on raking/calibration. Raking is used if it can be used (i.e., no numeric adjustment variables and always.calibrate not FALSE)

Usage

Calibrate(
  categorical.variables = NULL,
  categorical.targets = NULL,
  numeric.variables = NULL,
  numeric.targets = NULL,
  lower = NA,
  upper = NA,
  trim.iterations = 20,
  package = c("CVXR", "survey", "icarus")[1],
  always.calibrate = FALSE,
  subset = NULL,
  input.weight = NULL
)

Arguments

categorical.variables	An optional list or data frame of categorical adjustment variables.
categorical.targets	The target probabilities for each category listed in categorical.variables
numeric.variables	An optional list or data frame of categorical adjustment variables.
numeric.targets	the target mean for each numeric variable in numeric.variables
lower	A lower bound weight value (not guaranteed to be achieved).
upper	An upper bound weight value (not guaranteed to be achieved).
trim.iterations	The number of times to run the trim loop over the final weightings
package	The R package used to calibrate the model when raking is not conducted. Defaults to CVXR (see https://cvxr.rbind.io/cvxr_examples/cvxr_survey_calibration/). Other options are icarus and survey. . Use icarus with care, as sometimes target categories can be switched around
always.calibrate	If FALSE, which is the default, problems with only categorical adjustment variables are solved via iterative-proportional fitting (raking). Otherwise, they are solved via calibration.
subset	A logical vector indicating which subset of cases should be used to create the weight
input.weight	An optional weight variable; if supplied, the created weight is created to be as close to this input.weight as possible

Value

numeric A vector of weights

Displayr/flipData documentation built on March 2, 2024, 3:49 a.m.