run_benchmarking: Run Formal Benchmarking

In senseweight: Sensitivity Analysis for Weighted Estimators

Run Formal Benchmarking

Description

Wrapper function to run formal benchmarking on a set of pre-specified covariates. Returns a data.frame containing the benchmarked parameter values, the estimated bias, MRCS, and minimum k_sigma and k_rho values for a killer confounder.

Usage

run_benchmarking(
  estimate,
  RV,
  formula = NULL,
  weights = NULL,
  pop_svy = NULL,
  sample_svy = NULL,
  Y = NULL,
  weighting_vars = NULL,
  benchmark_vars = "all",
  data = NULL,
  treatment = NULL,
  outcome = NULL,
  selection = NULL,
  population_targets = NULL,
  weighting_method = "ebal",
  weight_max = Inf,
  sigma2 = NULL,
  estimand = "ATT"
)

Arguments

estimate	Weighted estimate
RV	Robustness Value
formula	Raking formula for survey estimand
weights	A vector, containing the estimated survey weights
pop_svy	Survey object, containing the population the survey sample is being re-weighted to
sample_svy	Survey object, containing the survey sample being re-weighted
Y	outcome of interest (used for survey object)
weighting_vars	Vector of variables to use in the weights estimation for ATT or PATE
benchmark_vars	Vector of variables to benchmark parameters for. If benchmark_vars = 'all', benchmarking will be run across all variables included in the weights. If not set to all, benchmarking will be conducted across the covariates included in the vector.
data	A data.frame containing the observed covariates included in the weights; must include variables specified in weighting_vars
treatment	Denotes which variable is the treatment variable
outcome	Denotes which variable is the outcome variable
selection	Denotes which variable is the selection variable
population_targets	Population targets for the raking formula in surveys (optional, if not provided, will be generated from pop_svy)
weighting_method	Weighting method. Supports weighting methods from the package WeightIt.
weight_max	Maximum weight to trim at. Default set to Inf.
sigma2	If estimand = "PATE", sigma2 must specify the bound on treatment effect heterogeneity. For the other two estimands, the function will automatically calculate the sample variance across the control units, or the survey sample.
estimand	Specifies estimand; possible parameters include "ATT", "PATE", or "Survey"

Value

data.frame containing the benchmarked parameter values, the estimated bias, MRCS, and minimum k_sigma and k_rho values for a killer confounder for the set of pre-specified covariates.

Examples

# For the external validity setting: 
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

# Select weighting variables:
weighting_vars <- names(df_all)[which(!names(df_all) %in% c("site", "S", "Y", "T"))]

# Run benchmarking:
df_benchmark <- run_benchmarking(
  weighting_vars = weighting_vars,
  data = df_all[, -1],
  treatment = "T", outcome = "Y", selection = "S",
  estimate = ipw,
  RV = RV, sigma2 = vartau,
  estimand = "PATE"
)

print(df_benchmark)

senseweight documentation built on Aug. 23, 2025, 1:11 a.m.