Simulate and Analyze Interval- and Mixed-Censored Survival Data.

simIC

The simIC package provides tools for simulating and analyzing interval-censored survival data, including left-, right-, and uncensored observations, using a variety of parametric distributions. It is useful for teaching, model development, and method evaluation in survival analysis.

✨ Features

• Supports commonly used parametric distributions:

  • Weibull
  • Exponential
  • Log-Normal
  • Logistic
  • Normal
  • Log-Logistic
  • Gamma
  • Gompertz
  • EMV (Extreme Minimum Value / Gumbel)

• Simulates survival data with interval, left, right, and uncensored observations using user-defined visit schedules (start_time, end_time) and an optional tolerance (uncensored_tol) for detecting exact event times.

• Provides two estimation functions:

  • mle_int()

    • Performs direct maximum likelihood estimation
    • Automatically detects and handles:
      • Interval-censored: contribution from F(Ri) - F(Li)
      • Left-censored: contribution from F(Ri)
      • Right-censored: contribution from 1 - F(Li)
      • Uncensored: contribution from f(ti)

  • mle_imp()

    • Uses imputation-based likelihood for interval- and uncensored data
    • For left- and right-censored, uses the proper likelihood contributions without imputation
    • Specifically:
      • Interval-censored: event times imputed from (Li, Ri) using midpoint, random, medians, or survival-based methods
      • Left-censored: contribution from F(Ri)
      • Right-censored: contribution from 1 - F(Li)
      • Uncensored: contribution from f(ti)

📦 Installation

You can install the development version of simIC from GitHub:

install.packages("remotes")
remotes::install_github("jayarasan/simIC")
library(simIC)

🧪 Simulate Survival Data
# Interval-censored data only (no visit window)
data <- simIC(n = 100, dist = "weibull", shape = 1.5, scale = 5, width = 2)

# Left-, right-, and uncensored data using a follow-up window and tolerance
data <- simIC(n = 100, dist = "weibull", shape = 1.5, scale = 5,
              width = 2, start_time = 0, end_time = 10, uncensored_tol = 0.1)

📈 Model Fitting Examples

# Direct MLE for interval-censored data
fit_int <- mle_int(data$left, data$right, dist = "weibull")
print(fit_int$estimates)

# Imputation-based MLE (midpoint)
fit_imp <- mle_imp(data$left, data$right, dist = "weibull", impute = "midpoint")
print(fit_imp$estimates)