Bayesian Treatment Effect Weight Estimation for Censored Data

This function estimates Bayesian importance sampling weights for treatment models and censoring models across multiple time points via JAGS

Usage

bayesweight_cen(
  trtmodel.list,
  cenmodel.list,
  data,
  n.chains = 2,
  n.iter = 25000,
  n.burnin = 15000,
  n.thin = 5,
  seed = NULL,
  parallel = TRUE
)

Arguments

trtmodel.list: A list of formulas corresponding to each time point with the time-specific treatment variable on the left-hand side and pre-treatment covariates to be balanced on the right-hand side. The formulas must be in temporal order, and must contain all covariates to be balanced at that time point. Interactions and functions of covariates are allowed.
cenmodel.list: A list of formulas for the censored data at each time point, with censoring indicators on the left-hand side and covariates on the right-hand side. The formulas must be in temporal order, and must contain all covariates to be balanced at that time point.
data: A data set in the form of a data frame containing the variables in "trtmodel.list" and "cenmodel.list". This must be a wide data set with exactly one row per unit.
n.chains: Integer specifying the number of MCMC chains to run. Set to 1 for non-parallel computation. For parallel computation, it is required to use at least 2 chains. The default is 2.
n.iter: Integer specifying the total number of iterations for each chain (including burn-in). The default is 25000.
n.burnin: Integer specifying the number of burn-in iterations for each chain. The default is 15000.
n.thin: Integer specifying the thinning rate for the MCMC sampler. The default is 5.
seed: Starting seed for the JAGS model. The default is NULL.
parallel: Logical scalar indicating whether to run the MCMC chains in parallel. The default is TRUE.

Value

A list of the calculated weights and the JAGS model where "weights" is a vector of posterior mean weights, computed by taking the average of the weights across all MCMC iterations and `model_string` is a character of the JAGS model based on the input of "trtmodel.list".

Examples

amodel <- list(
 c("(Intercept)" = -0.3, "L1_1" = 0.4, "L2_1" = -0.2),
 c("(Intercept)" = -0.1, "L1_2" = 0.3, "L2_2" = -0.1, "A_prev" = 0.5))
ymodel <- c(
"(Intercept)" = -0.8,
"A1"          = 0.2,
"A2"          = 0.4,
"L1_2"        = 0.3,
"L2_2"        = -0.3)
cmodel <- list(
 c("(Intercept)" = -1.5, "L1_1" = 0.2, "L2_1" = -0.2, "A" = 0.2),
 c("(Intercept)" = -1.5, "L1_2" = 0.1, "L2_2" = -0.1, "A" = 0.3))
testdata <- simData(
 n                = 50,
 n_visits         = 2,
 covariate_counts = c(2, 2),
 amodel           = amodel,
 ymodel           = ymodel,
 y_type           = "binary",
 right_censor     = TRUE,
 cmodel           = cmodel,
 seed             = 123
)
weights_cen <- bayesweight_cen(
                trtmodel.list = list(
                       A1 ~ L1_1 + L2_1,
                       A2 ~ L2_2 + L2_2 + A1),
                cenmodel.list = list(
                C1 ~ L1_1 + L2_1 + A1,
                C2 ~ L1_2 + L2_2 + A2),
                data = testdata,
                n.chains = 1,
                n.iter = 20,
                n.burnin = 10,
                n.thin = 1,
                seed = 890123,
                parallel = FALSE)
#> Compiling model graph
#>    Resolving undeclared variables
#>    Allocating nodes
#> Graph information:
#>    Observed stochastic nodes: 360
#>    Unobserved stochastic nodes: 20
#>    Total graph size: 1264
#> 
#> Initializing model
#> 
#> Warning: Some parameters have not converged with Geweke index > 1.96. More iterations may be needed.
summary(weights_cen)
#>              Length Class  Mode     
#> weights      50     -none- numeric  
#> model_string  1     -none- character