Title: Simulate Survival Data
Version: 1.0.0
Date: 2025-11-05
Description: Provides tools for simulating synthetic survival data using a variety of methods, including kernel density estimation, parametric distribution fitting, and bootstrap resampling techniques for a desired sample size.
Encoding: UTF-8
RoxygenNote: 7.3.2
Suggests: testthat (≥ 3.0.0)
Config/testthat/edition: 3
Imports: fitdistrplus, kdensity, survival, univariateML, actuar, flexsurv, stats, FAdist
Depends: R (≥ 3.5)
LazyData: true
License: MIT + file LICENSE
NeedsCompilation: no
Packaged: 2025-12-02 10:42:34 UTC; thurow
Author: Maria Thurow ORCID iD [aut, cre], Manasi Butee [aut], Ina Dormuth [ctb], Christina Sauer [ctb], Marc Ditzhaus [ctb], Markus Pauly [ctb]
Maintainer: Maria Thurow <maria.thurow@tu-dortmund.de>
Repository: CRAN
Date/Publication: 2025-12-09 07:50:18 UTC

Simulate Datasets Using Various Simulation Models

Description

Simulates survival datasets(Time-to-event data) based on original or reconstructed data using four different simulation models: Kernel Density Estimation (KDE), parametric distributions, conditional bootstrap, and Case Resampling. This function is designed to support comprehensive survival analysis simulations.

Usage

RealSurvSim(
  dat,
  col_time,
  col_status,
  col_group,
  reps = 10000,
  random_seed = 123,
  n = NULL,
  simul_type = c("cond", "case", "distr", "KDE"),
  distribs = c("exp", "exp", "exp", "exp")
)

Arguments

dat

A data.frame representing the original or reconstructed dataset for simulation. The dataset must include three columns: for event times, for censoring status, and for group identifiers.

col_time

The name or index of the column in dat representing time to event.

col_status

The name or index of the column in dat representing the event status (1 for event occurred, 0 for censored).

col_group

The name or index of the column in dat representing group assignments.

reps

The number of iterations, equivalent to the number of datasets simulated for each simulation model. Defaults to 10000.

random_seed

Seed for random number generation to ensure reproducibility. Defaults to 123.

n

An optional numeric vector specifying the number of observations to simulate for each group. If NULL, the function uses the original dataset's group sizes for simulation. For all simulation types except "conditional bootstrap," n can be set to arbitrary values, such as c(50, 60), where each element specifies the number of observations for a group. Defaults to NULL.

simul_type

A vector of characters specifying the types of simulation to perform. It includes "cond" (conditional bootstrap), "case" (case resampling), "distr" (parametric distributions), and "KDE" (kernel density estimation, supports all kernels available in the kdensity function. Refer to 'kdensity'). Note: Only one simulation type can be used at a time.

distribs

Character vector of length 4, one distribution per stratum. Must be one of:

  • "inverse_gamma"

  • "llogis"

  • "gumbel"

  • "exp"

  • "gamma"

  • "normal"

  • "cauchy"

Defaults to c("exp", "exp", "exp", "exp").

Value

A list containing the simulated datasets for each specified simulation model. The structure of the output list is as follows: 

     - {datasets}: A list of data frames, where each data frame represents a simulated dataset.
     - Each data frame contains:
         - {V1}: A numeric vector representing the simulated time-to-event data.
         - {V2}: A numeric or integer vector indicating the status, representing
           whether the event of interest has occurred (1) or is censored (0).
         - {V3}: An integer vector representing group.

     - The number of data frames within {datasets} corresponds to the number of repetitions specified
       by the {reps} parameter.

Examples

# liang should have columns: V1 (time), V2 (status), V3 (group)

         # Simulate data using parametric distribution fitting

          liang<- dats$Liang
          liang_distr <- RealSurvSim(
           dat = liang,
           col_time = "V1",
           col_status = "V2",
           col_group = "V3",
           reps = 10,
           simul_type = "distr",
           distribs = c("exp", "exp", "exp", "exp")
         )

Simulate Data Using Bootstrap Methods

Description

Simulates event and censoring times from an original dataset using specified bootstrap methodologies. This function supports conditional and case resampling bootstrap methods, allowing for flexible data simulation scenarios tailored to survival analysis.

Usage

data_simul_Bootstr(dat, n = NULL, type = "cond")

Arguments

dat

A dataframe containing the original dataset, expected to include columns for event times (V1), censoring indicators (V2), and group indicators (optional).

n

Integer specifying the number of observations to simulate. If NULL, the function simulates the same number of observations as in the original dataset. Defaults to NULL.

type

Character string specifying the type of bootstrap method to be used. Supported types include "cond" for conditional and "case" for case resampling. Defaults to "cond".

Value

A dataframe or a numeric vector of simulated values depending on the chosen bootstrap method. - For "case" bootstrap - For "cond" bootstrap, the arbitary n function does not work

Examples

dat <- data.frame(
  V1 = rexp(100, rate = 0.1), # Time-to-event data
  V2 = sample(0:1, 100, replace = TRUE),
  V3 = sample(0:1, 100, replace = TRUE)# Event indicator (0 = censored, 1 = event)
)
simulated_case <- data_simul_Bootstr(dat = dat, n = 100, type = "case")
simulated_cond <- data_simul_Bootstr(dat = dat, type = "cond")

Simulate Data Based on Parametric Distribution Estimates

Description

This function simulates data based on parameter estimates from a specified parametric distribution. It fits a chosen distribution to the original dataset and samples new values from this fitted distribution. Supported distributions include "inverse_gamma", "llogis" (log-logistic), "gumbel". "log-normal", "gamma", "exp", "cauchy".

Usage

data_simul_Estim(orig_vals, n = NULL, distrib = "exp")

Arguments

orig_vals

Numeric vector of values from the original dataset.

n

Integer specifying the number of observations to simulate. If NULL, the function simulates the same number of observations as in the original dataset. Defaults to NULL.

distrib

Character; one of "inverse_gamma", "llogis", "gumbel", "exp", "gamma", "normal", or "cauchy".

Value

Numeric vector of n simulated values based on the fitted parametric distribution.

Examples

original_data <- rnorm(100, mean = 50, sd = 10)
simulated_data <- data_simul_Estim(orig_vals = original_data, n = 100, distrib = "inverse_gamma")

Kernel Density Estimation-based Data Simulation

Description

Simulates data based on the kernel density estimation (KDE) of given data. KDE is a non-parametric way to estimate the probability density function of a random variable. This function applies the accept-reject method to generate values that follow the estimated density of the original dataset.

Usage

data_simul_KDE(orig_vals, n = NULL, kernel = "gaussian")

Arguments

orig_vals

Numeric vector of values from the original dataset.

n

Integer, number of observations to simulate. If NULL, the function simulates the same number of observations as in the original dataset. Defaults to NULL.

kernel

Character, specifying the kernel to be used for KDE. Defaults to "gaussian".

Value

Numeric vector of n simulated values.

Examples

original_data <- c(rnorm(100, mean = 50, sd = 10))
simulated_data <- data_simul_KDE(original_data, n = 100)

Collection of Survival Datasets (dats)

Description

dats is a collection of seven survival datasets used for testing and simulation of survival data. These datasets were reconstructed from published Kaplan-Meier survival curves using the widely applied algorithm by Guyot et al. (2012). The datasets were originally sourced from various clinical studies and digitized using WebPlotDigitizer. They are used as benchmarks for synthetic survival data methods, including kernel density estimation, parametric distribution fitting, and bootstrap resampling.

Usage

data(dats)

Format

A list containing 7 data frames. Each data frame includes:

V1

Time to event (numeric).

V2

Event indicator (0 = censored, 1 = event; numeric).

V3

Group identifier (numeric or categorical).

The datasets in dats are:

Source

Examples

data(dats)
names(dats)
head(dats$Liang)

Map Distribution Names for Density and Random Generation Functions

Description

This function maps a user-specified distribution name to a list containing the names of the corresponding density and random generation functions. If the distribution is not explicitly mapped, it assumes the standard naming convention by prepending "d" or "r" to the distribution name.

Usage

get_dist_names(distrib)

Arguments

distrib

A character string specifying the distribution.

Value

A list with components d and r, which are character strings representing the names of the density and random generation functions, respectively.

See Also

get_fitdist_name

Map Distribution Name for Fitting

Description

This function maps a user-specified distribution name to the appropriate name used for parameter estimation with fitdist() from the fitdistrplus package.

Usage

get_fitdist_name(distrib)

Arguments

distrib

A character string specifying the distribution (e.g., "poisson", "exp", "gumbel").

Value

A character string representing the distribution name as expected by fitdist().

See Also

get_dist_names