Simulate single-event survival data with informative censoring

Generates a dataset where event times follow a Weibull distribution parameterized by a binary treatment covariate x, and censoring depends on an ancillary biomarker W2 through a shared gamma frailty, inducing informative censoring using Clayton's copula. This is the data-generating mechanism used in the single-event guided example.

Usage

sim_data_se(
  n = 500,
  alpha = 0.05,
  x_prop = 0.5,
  a = 2,
  sigma = 500,
  beta = log(0.25),
  lambda = 0.01,
  phi = -5
)

Arguments

n

Integer. Number of subjects to simulate. Default is 500.

alpha

Numeric. Shape parameter for the gamma frailty distribution, which has a fixed rate of 1. Here, alpha denotes the level of dependence between the survival time and the covariate that drives censoring. A smaller alpha leads to a stronger dependence. This parameter is also related to Kendall's correlation coefficient, \(\tau\), such that \(\tau = (1/\alpha)/(1/\alpha + 2)\). Default is 0.05.

x_prop

Numeric in (0, 1). Probability of treatment (x = 1). Default is 0.5.

a

Numeric. Weibull shape parameter for the event time distribution. Default is 2.

sigma

Numeric. Weibull scale parameter for the event time distribution. Default is 500.

beta

Numeric. Log hazard ratio for the treatment effect on event time. Default is log(0.25).

lambda

Numeric. Baseline censoring rate when W2=0. Default is 0.01.

phi

Numeric. The log hazard ratio for the association between a one-unit increase in W2 and censoring time. Negative values mean higher W2 leads to a lower censoring rate (i.e., longer follow-up for high-W2 subjects). Default is -5.

Value

A data frame with columns:

S

True (latent) event time.

t

Observed time (minimum of S and the censoring time).

delta

Event indicator: 1 = event occurred, 0 = censored.

x

Binary treatment indicator (0 or 1).

W2

Ancillary biomarker covariate that drives informative censoring.

Examples

set.seed(20240429)
dat <- sim_data_se(n = 500)
table(dat$delta)
#> 
#>   0   1 
#> 198 302 
summary(dat$t)
#>      Min.   1st Qu.    Median      Mean   3rd Qu.      Max. 
#>    0.6884  172.7074  441.1759  552.9262  803.2563 2429.5429