Estimates causal effects using either Inverse Probability Weighting (IPW) or Augmented Inverse Probability Weighting (AIPW) methods.
Usage
ipw_estimators(
data,
estimation_method,
outcome,
treatment,
covariates,
model_specification = "linear",
output = FALSE,
...
)Arguments
- data
Data frame containing the analysis dataset
- estimation_method
Character, either 'IPW' or 'AIPW'
- outcome
Character, name of the outcome variable
- treatment
Character, name of the treatment variable
- covariates
Character vector of covariate names
- model_specification
Character, type of model specification ('linear', 'interaction', or 'splines')
- output
Logical, whether to return detailed output
- ...
Additional arguments passed to underlying functions: For IPW method: cv.glmnet() arguments: All arguments are supported with their default values.
For AIPW method: causal_forest() arguments: All arguments are supported with num.trees defaulting to 100 instead of 2000. Other arguments maintain their default values.
average_treatment_effect() arguments: All arguments are supported with method defaulting to 'AIPW'. Other arguments maintain their default values.
Value
If output=TRUE, returns a list containing:
- estimation_value
The estimated treatment effect
- ate_results
Detailed results including standard errors and p-values
- e_hat
Estimated propensity scores
- model_spec_matrix
Model specification matrix
- treatment_variable
Treatment indicator vector
Examples
if (FALSE) { # \dontrun{
data <- data.frame(
y = rnorm(100),
w = rbinom(100, 1, 0.5),
x1 = rnorm(100),
x2 = rnorm(100)
)
# Basic usage
results <- ipw_estimators(
data = data,
estimation_method = "IPW",
outcome = "y",
treatment = "w",
covariates = c("x1", "x2"),
model_specification = 'linear',
output = TRUE
)
# Using additional parameters with AIPW
results_aipw <- ipw_estimators(
data = data,
estimation_method = "AIPW",
outcome = "y",
treatment = "w",
covariates = c("x1", "x2"),
model_specification = 'linear',
output = TRUE,
# causal_forest() arguments
num.trees = 100,
# average_treatment_effect() arguments
target.sample = "control"
)
} # }