Package 'RealSurvSim'

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

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")

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

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")

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

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)

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:

• Liang: Derived from Liang et al. (2019).

• Spigel: Derived from Spigel et al. (2022).

• Wu: Derived from Wu et al. (2015).

• Wei: Derived from Wei et al. (2020).

• Lima: Derived from Lima et al. (2018).

• Yoshioka: Derived from Yoshioka et al. (2019).

• Seto: Derived from Seto et al. (2020).

Source

• Maria Thurow et al. (2024). "How to Simulate Realistic Survival Data? A Simulation Study to Compare Realistic Simulation Models" arXiv preprint, https://arxiv.org/abs/2308.07842.

• Original datasets from respective publications (see dataset documentation for details).

• Data reconstructed using the algorithm by Guyot et al. (2012), BMC Medical Research Methodology, doi:10.1186/1471-2288-12-9.

• Data digitized using WebPlotDigitizer (Rohatgi, A.), https://automeris.io/WebPlotDigitizer.

Examples

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

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

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")
         )