dot-weightit_methods: Weighting methods
In WeightIt: Weighting for Covariate Balance in Observational Studies

.weightit_methods

R Documentation

Weighting methods

Description

.weightit_methods is a list containing the allowable weighting methods that can be supplied by name to the method argument of weightit(), weightitMSM(), and weightit.fit(). Each entry corresponds to an allowed method and contains information about what options are and are not allowed for each method. While this list is primarily for internal use by checking functions in WeightIt, it might be of use for package authors that want to support different weighting methods.

Usage

.weightit_methods

Format

An object of class list of length 10.

Details

Each component is itself a list containing the following components:

treat_type: at least one of "binary", "multi-category", or "continuous" indicating which treatment types are available for this method.
estimand: which estimands are available for this method. All methods that support binary and multi-category treatments accept "ATE", "ATT", and "ATC", as well as some other estimands depending on the method. See get_w_from_ps() for more details about what each estimand means.
alias: a character vector of aliases for the method. When an alias is supplied, the corresponding method will still be dispatched. For example, the canonical method to request entropy balancing is "ebal", but "ebalance" and "entropy" also work. The first value is the canonical name.
description: a string containing the description of the name in English.
ps: a logical for whether propensity scores are returned by the method for binary treatments. Propensity scores are never returned for multi-category or continuous treatments.
msm_valid: a logical for whether the method can be validly used with longitudinal treatments.
msm_method_available: a logical for whether a version of the method can be used that estimates weights using a single model rather than multiplying the weights across time points. This is related to the is.MSM.method argument of weightitMSM().
subclass_ok: a logical for whether subclass can be supplied to compute subclassification weights from the propensity scores.
packages_needed: a character vector of the minimal packages required to use the method. Some methods may require additional packages for certain options.
s.weights_ok: a logical for whether sampling weights can be used with the method.
missing: a character vector of the allowed options that can be supplied to missing when missing data is present. All methods accept "ind" for the missingness indicator approach; some other methods accept additional values.
moments_int_ok: a logical for whether moments, int, and quantile can be used with the method.
moments_default: when moments_int_ok is TRUE, the default value of moments used with the method. For most methods, this is 1.
density_ok: a logical for whether arguments that control the density can be used with the method when used with a continuous treatment.
stabilize_ok: a logical for whether the stabilize argument (and num.formula for longitudinal treatments) can be used with the method.
plot.weightit_ok: a logical for whether plot() can be used on the weightit output with the method.

Examples

# Get all acceptable names
names(.weightit_methods)

# Get all acceptable names and aliases
lapply(.weightit_methods, `[[`, "alias")

# Which estimands are allowed with `method = "bart"`
.weightit_methods[["bart"]]$estimand

# All methods that support continuous treatments
supp <- sapply(.weightit_methods, function(x) {
  "continuous" %in% x$treat_type
})
names(.weightit_methods)[supp]

# All methods that return propensity scores (for
# binary treatments only)
supp <- sapply(.weightit_methods, `[[`, "ps")
names(.weightit_methods)[supp]

WeightIt documentation built on April 4, 2025, 2:05 a.m.