generalize: Generalize Average Treatment Effect from Randomized Trial to...

In benjamin-ackerman/generalize: Generalizing Average Treatment Effects from RCTs to Target Populations

Description

Generalize Average Treatment Effect from Randomized Trial to Population

Usage

generalize(
  outcome,
  treatment,
  trial,
  selection_covariates,
  data,
  method = "weighting",
  selection_method = "lr",
  sl_library = NULL,
  survey_weights = FALSE,
  trim_weights = FALSE,
  trim_pctile = 0.97,
  is_data_disjoint = TRUE,
  trimpop = FALSE,
  seed
)

Arguments

outcome	variable name denoting outcome
treatment	variable name denoting binary treatment assignment (ok if only available in trial, not population)
trial	variable name denoting binary trial participation (1 = trial participant, 0 = not trial participant)
selection_covariates	vector of covariate names in data set that predict trial participation
data	data frame comprised of "stacked" trial and target population data
method	method to generalize average treatment effect to the target population. Default is "weighting" (weighting by participation probability). Other methods supported are "BART" (Bayesian Additive Regression Trees - NOT READY YET) and "TMLE" (Targeted Maximum Likelihood Estimation)
selection_method	method to estimate the probability of trial participation. Default is logistic regression ("lr"). Other methods supported are Random Forests ("rf") and Lasso ("lasso")
sl_library	vector of SuperLearner library methods. If ‘selection_method' = ’super', specify names of methods to include in library. Default is NULL.
survey_weights	variable name of population data's complex survey weights. Default is FALSE: if FALSE, then population data do not come a complex survey and weights do not need to be incorporated in estimation.
trim_weights	logical. If TRUE, then trim the weights to the value specified in 'trim_pctile'. Default is FALSE.
trim_pctile	numeric. If 'trim_weights' is TRUE, then specify what percentile weights should be trimmed to. Default is 0.97.
is_data_disjoint	logical. If TRUE, then trial and population data are considered independent. This affects calculation of the weights - see details for more information.
trimpop	logical. If TRUE, then population data are subset to exclude individuals with covariates outside bounds of trial covariates.
seed	numeric. By default, the seed is set to 13783, otherwise can be specified (such as for simulation purposes).

Value

generalize returns an object of the class "generalize"

benjamin-ackerman/generalize documentation built on Oct. 11, 2020, 3:58 a.m.