riAFTBART (version 0.3.2)

sa: Flexible Monte Carlo sensitivity analysis for unmeasured confounding

Description

This function implements the flexible sensitivity analysis approach for unmeasured confounding with multiple treatments from multilevel survival data.

Usage

sa(
  M.burnin,
  M.keep,
  M.thin = 1,
  status,
  y.train,
  x.train,
  trt.train,
  x.test,
  trt.test,
  cluster.id,
  verbose = FALSE,
  formula = NULL,
  prior_c_function,
  Q1,
  Q2 = NULL,
  nCores = 1,
  ...
)

Arguments

M.burnin

A numeric value indicating the number of MCMC iterations to be treated as burn in.

M.keep

A numeric value indicating the number of MCMC posterior draws after burn in.

M.thin

A numeric value indicating the thinning parameter.

status

A vector of event indicators: status = 1 indicates that the event was observed while status = 0 indicates the observation was right-censored.

y.train

A vector of follow-up times.

x.train

A dataframe or matrix, including all the covariates but not treatments for training data, with rows corresponding to observations and columns to variables.

trt.train

A numeric vector representing the treatment groups for the training data.

x.test

A dataframe, including all the covariates but not treatments for testing data, with rows corresponding to observations and columns to variables.

trt.test

A numeric vector representing the treatment groups for the testing data.

cluster.id

A vector of integers representing the clustering id.

verbose

A logical indicating whether to show the progress bar. The default is FALSE

formula

A formula object for the analysis. The default is to use all terms specified in x.train.

prior_c_function

1) A vector of characters indicating the prior distributions for the confounding functions. Each character contains the random number generation code from the standard probability distributions in the stats package. 2) A vector of characters including the grid specifications for the confounding functions. It should be used when users want to formulate the confounding functions as scalar values. 3) A matrix indicating the point mass prior for the confounding functions

Q1

A numeric value indicating the number of draws of the GPS from the posterior predictive distribution

Q2

A numeric value indicating the number of draws from the prior distributions of the confounding functions

nCores

A numeric value indicating number of cores to use for parallel computing.

...

Other parameters that can be passed to BART functions

Value

A list with the following elements:

result_riAFTBART:

Corrected log survival time for the test data from the riAFT-BART model.

c_functions:

The confounding functions sampled from the specified distribution used in the analysis.

Examples

# NOT RUN {
set.seed(20181223)
n = 5       # number of clusters
k = 50      # cluster size
N = n*k     # total sample size
cluster.id = rep(1:n, each=k)
tau.error = 0.8
b = rnorm(n, 0, tau.error)
alpha = 2
beta1 = 1
beta2 = -1
beta3 = -2
sig.error = 0.5
censoring.rate = 0.02
x1 = rnorm(N,0.5,1)
x2 = rnorm(N,1.5,0.5)
trt.train = sample(c(1,2,3), N, prob = c(0.4,0.3,0.2), replace = TRUE)
trt.test = sample(c(1,2,3), N, prob = c(0.3,0.4,0.2), replace = TRUE)
error = rnorm(N,0,sig.error)
logtime = alpha + beta1*x1 + beta2*x2 + b[cluster.id] + error
y = exp(logtime)
C = rexp(N, rate=censoring.rate) # censoring times
Y = pmin(y,C)
status = as.numeric(y<=C)
res_sa <- sa(M.burnin = 10, M.keep = 10, M.thin = 1, status = status,
             y.train = Y,trt.train = trt.train,trt.test = trt.test,
             x.train = cbind(x1,x2),
             x.test = cbind(x1,x2),
             cluster.id = cluster.id, verbose = F,prior_c_function = c(
               "runif(-0.6, 0)",# c(1,2)
               "runif(0, 0.6)",# c(2,1)
               "runif(-0.6, 0)", # c(2,3)
               "seq(-0.6, 0, by = 0.3)", # c(1,3)
               "seq(0, 0.6, by = 0.3)", # c(3,1)
              "runif(0, 0.6)" # c(3,2)
            ),Q1 = 1, nCores = 1)
 
# }