get_variance_estimation: Calculate the variance (SE) of the marginal treatment effects...

In bunsen: Marginal Survival Estimation with Covariate Adjustment

Calculate the variance (SE) of the marginal treatment effects (hazard ratio) adjusting covariates in clinical trials

Description

Estimate the standard error or variance of the marginal treatment effects using nonparametric bootstrap. Currently, this only supports clustermq for parallel computation.

Usage

get_variance_estimation(
  cox_event,
  cox_censor,
  trt,
  data,
  M,
  n.boot,
  seed = NULL,
  cpp = TRUE,
  control = clmqControl(),
  verbose = TRUE
)

Arguments

cox_event	Object. A coxph model using the survival time and survival status.
cox_censor	Object. A coxph model using the survival time and 1-survival status.
trt	Character. Variable name of the treatment assignment. Only support two arm trial at the moment.
data	A data frame used for cox_event and cox_censor.
M	Numeric. The number of simulated counterfactual patients. Suggest to set above 1,000,000 to get robust estimation but it is time comsuming,
n.boot	Numeric. Number of bootstrap.
seed	Numeric. Random seed for simulation.
cpp	Bool. True for using C++ optimization. False for not using C++ optimization. This requires cpp package installed.
control	Named list. A list containing control parameters, including memory of remote workers, whether to use nested parallel computation or local multiprocess, number of remote workers/jobs, etc. See details of clmqControl.
verbose	Bool. Print status messages. Default: TRUE

Details

If clustermq is not available, we suggest building your own bootstrap like boot and doParallel by using the function – get_point_estimate. This can also get you the SE or variance estimates. If you only run this function, you need to have cox_censor and cox_event in the environment.

Value

A vector containing SE and 95% CI.

References

Daniel R, Zhang J, Farewell D. Making apples from oranges: Comparing noncollapsible effect estimators and their standard errors after adjustment for different covariate sets. Biom J. 2021;63(3):528-557. doi:10.1002/bimj.201900297

Examples

## Not run: 
#Don't run as it requires LSF scheduler
library(survival)
data("oak")

cox_event <- coxph(Surv(OS, os.status) ~ trt + btmb + pdl1, data = oak)
#
cox_censor <- coxph(Surv(OS, 1 - os.status) ~ trt + btmb + pdl1, data = oak)
#
get_variance_estimation(cox_event, cox_censor,
  trt = "trt", data = oak,
  M = 1000, n.boot = 10, control = clmqControl(), cpp = FALSE
)

## End(Not run)

bunsen documentation built on June 8, 2025, 11:34 a.m.