Title: Simulation and Analysis of Event History Data
Version: 0.1.1
Description: Simulate event history data from a framework where treatment decisions and disease progression are represented as counting process. The user can specify number of events and parameters of intensities thereby creating a flexible simulation framework.
License: MIT + file LICENSE
Encoding: UTF-8
RoxygenNote: 7.3.2
URL: https://github.com/miclukacova/simevent
BugReports: https://github.com/miclukacova/simevent/issues
Imports: stats, data.table, ggplot2, survival, Rcpp
Suggests: knitr, rmarkdown, testthat (≥ 3.0.0)
Config/testthat/edition: 3
LinkingTo: Rcpp
VignetteBuilder: knitr
NeedsCompilation: yes
Packaged: 2026-04-24 08:10:18 UTC; dns525
Author: Michaela Lukacova [aut, cre]
Maintainer: Michaela Lukacova <michaela.lukacova@sund.ku.dk>
Repository: CRAN
Date/Publication: 2026-04-24 08:30:02 UTC

simevent: Simulation and Analysis of Event History Data

Description

Simulate event history data from a framework where treatment decisions and disease progression are represented as counting process. The user can specify number of events and parameters of intensities thereby creating a flexible simulation framework.

Author(s)

Maintainer: Michaela Lukacova michaela.lukacova@sund.ku.dk

See Also

Useful links:

Transform Event Data into Interval Format for Classical Inference

Description

Converts simulated event history data with time-dependent covariates into an interval (start-stop) format, suitable for classical survival analysis functions like coxph. Adds interval start and stop times (tstart, tstop) and a counting variable k indexing events. Optionally, the function can split intervals at a specified time point to accomadate estimation of time-varying effects.

Usage

IntFormatData(data, N_cols = 6:9, timeVar = FALSE, t_prime = NULL)

Arguments

data

A data.frame or data.table containing event data with columns ID, Time, Delta, and counting process columns specified by N_cols.

N_cols

Integer vector. Column indices of data that correspond to counting process variables. Defaults to 6:9.

timeVar

Logical. If TRUE, the intervals are split at t_prime to allow time-varying covariate effects. Default is FALSE.

t_prime

Numeric scalar. Time point at which to split intervals if timeVar = TRUE.

Value

A data.table with columns tstart, tstop, k, and other original variables, formatted for survival analysis.

Examples

data <- simEventData(10)
IntFormatData(data)

Simulation and Estimation with Modified Shape Parameter

Description

This function simulates event history data from either the Disease, Treatment or Drop In setting. See simDisease, simTreatment and simDropIn. The shape parameter \eta of respectively the disease process, the Drop In process and the Treatment process is multiplied by alpha. The function either

Usage

alphaSim(
  N = 10000,
  eta = rep(0.1, 4),
  nu = rep(1.1, 4),
  alpha = 0.5,
  tau = 5,
  a0 = 1,
  years_lost = FALSE,
  setting = "Disease",
  return_data = FALSE,
  cens = 0,
  ...
)

Arguments

N

Integer. Number of individuals to simulate. Default is 10,000.

eta

Numeric vector. Shape parameters for Weibull hazards. Length of the vector should match number of events. For the Disease and Drop In setting this is 4. For the Treatment setting, this is 3. (default rep(0.1, 4)).

nu

Numeric vector. Scale parameters for Weibull hazards. Length of the vector should match number of events. For the Disease and Drop In setting this is 4. For the Treatment setting, this is 3. (default rep(1.1, 4)).

alpha

Numeric scalar. Multiplicative factor applied to the disease process shape parameter \eta.

tau

Numeric scalar. Time horizon at which proportions are computed.

a0

Binary (0/1). Specifies the group for comparison in setting Drop In and Disease.

years_lost

Logical. If TRUE, computes years lost instead of proportions.

setting

Character string. Must be either "Disease", "Drop In" or "Treatment". Depending on the simulation setting.

return_data

Logical. If TRUE the simulated data is returned.

cens

Binary scalar. Indicates whether individuals are at risk of censoring (default 0).

...

Additional arguments passed to respectively simDisease, simTreatment and simDropIn.

Value

A list with two components:

effect_L

Proportion (or years lost) of individuals diagnosed with disease by time \tau, under intervention.

effect_death

Proportion (or years lost) of individuals who died by time \tau under intervention.

Or the simulated data.

Examples

alphaSim(N = 100, eta = rep(0.1,3), nu = rep(1.1,3), alpha = 0.5, setting = "Disease")
alphaSim(N = 100, setting = "Drop In", beta_A0_Z = 1)

Estimate Effect of Intervention: Modifying Eta Parameter of Process

Description

This function simulates data from the disease setting in two scenarios. Under intervention on the shape parameter \eta of the disease process is multiplied by alpha, and a baseline (non-intervened) scenario. It computes the proportion of individuals who experience death or disease by a specified time \tau in the group A0 = a0, optionally returning years_lost. The function can also plot a sample of the event data for each scenario for comparison.

Usage

intEffectAlpha(
  N = 10000,
  setting = "Disease",
  eta = rep(0.1, 4),
  nu = rep(1.1, 4),
  alpha = 0.5,
  tau = 5,
  a0 = 1,
  years_lost = FALSE,
  plot = TRUE,
  lower = 10^(-30),
  upper = 200,
  cens = 0,
  ...
)

Arguments

N

Integer. Number of individuals to simulate. Default is 10,000.

setting

Character string. Must be either "Disease", "Drop In" or "Treatment". Depending on the simulation setting.

eta

Numeric vector of length 3. Shape parameters for Weibull hazards (default rep(0.1, 4)).

nu

Numeric vector of length 3. Scale parameters for Weibull hazards (default rep(1.1, 4)).

alpha

Numeric scalar. Multiplicative factor applied to the disease process shape parameter \eta.

tau

Numeric scalar. Time horizon at which proportions are computed.

a0

Binary (0/1). Specifies the group for comparison.Only relevant in setting "Drop In" and "Disease".

years_lost

Logical. If TRUE, computes years lost instead of proportions.

plot

Logical. If TRUE, plots timelines for sample of intervention and non intervention data.

lower

Numeric scalar. Lower bound for root-finding in hazard inversion (default 1e-30).

upper

Numeric scalar. Upper bound for root-finding in hazard inversion (default 200).

cens

Binary scalar. Indicates whether individuals are at risk of censoring (default 0).

...

Additional arguments passed to respectively simDisease, simTreatment and simDropIn.

Value

A list with two components:

effect_L

Proportion (or years lost) of individuals diagnosed with disease by time \tau in group A0 = a0, under intervention.

effect_death

Proportion (or years lost) of individuals who died by time \tau in group A0 = a0, under intervention.

Examples

intEffectAlpha(N = 1000, alpha = 0.7, tau = 5, years_lost = FALSE, a0 = 1, setting = "Drop In")

Wrapper for inverse cumulative hazard

Description

A wrapper around the Rcpp function inverseScHaz, used to find the inverse by numeric methods in case of no simple analytical solution.

Usage

inverseScHaz(
  p,
  t,
  lower,
  upper,
  eta,
  nu,
  phi,
  at_risk,
  tol = 1e-09,
  max_iter = 100
)

Arguments

p

The random variable (typically -log(U)).

t

Current time.

lower

Lower bound for root finding.

upper

Upper bound for root finding.

eta

Numeric vector of shape parameters.

nu

Numeric vector of scale parameters.

phi

Numeric vector of multiplicative effects.

at_risk

Numeric vector indicating at-risk indicators for each event type.

tol

Numeric tolerance for root-finding. Default is 1e-9.

max_iter

Maximum iterations. Default is 100.

Value

A numeric scalar, the root u.

Examples

eta <- c(0.1, 0.1)
nu <- c(1.1, 1.1)
phi <- c(1, 1)
at_risk <- c(1, 1)
inverseScHaz(0.5, t = 0, lower = 1e-15, upper = 200, eta, nu, phi, at_risk)

Wrapper for inverse cumulative hazard

Description

A wrapper around the Rcpp function inverseScHazCppTV, used to find the inverse of the summed cumulative hazard.

Usage

inverseScHazTV(
  p,
  t,
  lower = 1e-15,
  upper = 200,
  t_prime,
  eta,
  nu,
  phi,
  phi_prime,
  at_risk,
  tol = 1e-09,
  max_iter = 100
)

Arguments

p

The random variable (typically -log(U)).

t

The time of the previous event

lower

Lower bound for root finding.

upper

Upper bound for root finding.

t_prime

The time where the time varying effects change

eta

Numeric vector of shape parameters.

nu

Numeric vector of scale parameters.

phi

Numeric vector of multiplicative effect bedfore time t_prime

phi_prime

Numeric vector of multiplicative effects after time t_prime

at_risk

Numeric vector indicating at-risk indicators for each event type.

tol

Numeric tolerance for root-finding. Default is 1e-9.

max_iter

Maximum iterations. Default is 100.

Value

A numeric scalar, the root u.

Examples

eta <- c(0.1, 0.1)
nu <- c(1.1, 1.1)
phi <- c(1, 1)
at_risk <- c(1, 1)
phi_prime <- c(2, 2)
inverseScHazTV(p = 0.5, t= 1, t_prime = 2, eta = eta, nu = nu,
                       phi = phi, phi_prime = phi_prime, at_risk = at_risk)

Plot Simulated Event History Data

Description

Visualizes event history data by plotting individual event times colored and shaped by event type. Each individual's timeline is displayed horizontally with events marked along it.

Usage

plotEventData(data, title = "Event Data")

Arguments

data

A data.frame or data.table containing at least the columns ID, Time, and Delta.

title

Character string specifying the plot title. Defaults to "Event Data".

Value

A ggplot object representing the event data visualization.

Examples

data <- simEventData(10)
plotEventData(data)

Sample event types from matrix of probabilities

Description

Sample event types from matrix of probabilities

Usage

sampleEvents(probs)

Arguments

probs

A matrix where each column is a probability vector

Value

A vector of sampled event types (0-indexed)

Simulate Competing Risks Data

Description

Simulates competing risks data for N individuals who are at risk of mutually exclusive event types. Three event types are simulated, where one can be interpreted as censoring.

Usage

simCRdata(N, beta = NULL, eta = rep(0.1, 3), nu = rep(1.1, 3), cens = 1, ...)

Arguments

N

Integer. Number of individuals to simulate.

beta

Numeric matrix of dimension 2x3. Covariate effects of L0 and A0 on the three competing processes (columns correspond to processes). Defaults to zero matrix if NULL.

eta

Numeric vector of length 3. Shape parameters for Weibull hazards, parameterized as \eta \nu t^{\nu - 1}. Defaults to rep(0.1, 3).

nu

Numeric vector of length 3. Scale parameters for Weibull hazards. Defaults to rep(1.1, 3).

cens

Binary (0 or 1). Indicates if a censoring process is included. Default is 1.

...

Additional arguments passed to simEventData, including add_cov for extra covariates.

Details

The event intensities follow Weibull hazard models parameterized by shape and scale parameters \eta and \nu. Covariate effects on the hazard are specified by the beta matrix, which models the effects of baseline covariates L0 and A0 on each event type.

Value

A data.frame with simulated competing risk data including:

Examples

simCRdata(10)

Simulate Data in a Disease Setting

Description

This function simulates event data representing three event types: Censoring (0), Death (1), and Change in Covariate Process (2). Death and Censoring are terminal events, while Change in Covariate Process can occur only once.

Usage

simDisease(
  N,
  eta = rep(0.1, 3),
  nu = rep(1.1, 3),
  cens = 1,
  beta_L0_D = 1,
  beta_L0_L = 1,
  beta_L_D = 1,
  beta_A0_D = 0,
  beta_A0_L = 0,
  beta_L0_C = 0,
  beta_A0_C = 0,
  beta_L_C = 0,
  followup = Inf,
  lower = 10^(-15),
  upper = 200,
  beta_L_D_t_prime = NULL,
  t_prime = NULL,
  gen_A0 = NULL,
  at_risk_cov = NULL,
  ...
)

Arguments

N

Numeric scalar. Number of individuals to simulate.

eta

Numeric vector of length 3. Shape parameters for Weibull intensities with parameterization \eta \nu t^{\nu - 1}. Defaults to rep(0.1, 3).

nu

Numeric vector of length 3. Scale parameters for the Weibull hazards. Defaults to rep(1.1, 3).

cens

Binary scalar. Indicates whether individuals are at risk of censoring (default 1).

beta_L0_D

Numeric scalar. Effect of baseline covariate L0 on death risk (default 1).

beta_L0_L

Numeric scalar. Effect of baseline covariate L0 on covariate change risk (default 1).

beta_L_D

Numeric scalar. Effect of covariate change (L = 1) on death risk (default 1).

beta_A0_D

Numeric scalar. Effect of baseline treatment (A0 = 1) on death risk (default 0).

beta_A0_L

Numeric scalar. Effect of baseline treatment (A0 = 1) on covariate change risk (default 0).

beta_L0_C

Numeric scalar. Effect of baseline covariate L0 on censoring probability (default 0).

beta_A0_C

Numeric scalar. Effect of baseline treatment A0 on censoring probability (default 0).

beta_L_C

Numeric scalar. Effect of covariate change (L = 1) on censoring probability (default 0).

followup

Numeric scalar. Maximum follow-up (censoring) time. Defaults to Inf.

lower

Numeric scalar. Lower bound for root-finding (inverse cumulative hazard) (default 1e-15).

upper

Numeric scalar. Upper bound for root-finding (default 200).

beta_L_D_t_prime

Numeric scalar or NULL. Additional effect of covariate change on death risk after time t_prime (optional).

t_prime

Numeric scalar or NULL. Time point where effects change (optional).

gen_A0

Function. Function to generate the baseline treatment covariate A0. Takes N and L0 as inputs. Default is a Bernoulli(0.5) random variable.

at_risk_cov

Function. Function determining if an individual is at risk for each event type, given their covariates. Takes a numeric vector covariates and returns a binary vector. Default returns 1 for all events.

...

Additional arguments passed to simEventData or simEventTV

Details

Event intensities depend on previous events and predefined parameters \nu and \eta.

The arguments beta_X_Y control how the X affects Y. A positive value means that a higher value of X increases the intensity of Y, while a negative value decreases the intensity.

The simulation uses an event history framework with terminal events (death, censoring) and a single recurrent covariate change. The event intensities depend on covariates and previous events according to user-specified parameters. Time-varying effects can be included via beta_L_D_t_prime and t_prime.

Value

A data frame containing the simulated data with columns:

ID

Individual identifier

Time

Time of the event

Delta

Event type (0 = censoring, 1 = death, 2 = covariate change)

L0

Baseline covariate

L

Covariate indicating change in covariate process

Examples

simDisease(10)

simDropIn

Description

Simulate Event Data from a "Drop In" Setting

Usage

simDropIn(
  N,
  eta = c(0.5, 0.5, 0.1, 0.25),
  nu = c(1.1, 1.1, 1.1, 1.1),
  adherence = FALSE,
  followup = Inf,
  cens = 1,
  generate.A0 = function(N, L0) stats::rbinom(N, 1, 0.5),
  lower = 1e-200,
  upper = 1e+10,
  t_prime = NULL,
  at_risk_cov = NULL,
  beta_L_A = 1,
  beta_L_Z = 2,
  beta_L_D = 1.5,
  beta_L_C = 0,
  beta_A_L = -0.5,
  beta_A_Z = -0.5,
  beta_A_D = -1,
  beta_A_C = 0,
  beta_Z_L = -1,
  beta_Z_A = 0,
  beta_Z_D = -1,
  beta_Z_C = 0,
  beta_L0_L = 1,
  beta_L0_A = 1,
  beta_L0_Z = 1,
  beta_L0_D = 1,
  beta_L0_C = 0,
  beta_A0_L = -1.5,
  beta_A0_A = 0,
  beta_A0_Z = 0,
  beta_A0_D = -2,
  beta_A0_C = 0,
  beta_L_A_prime = 0,
  beta_L_Z_prime = 0,
  beta_L_D_prime = 0,
  beta_L_C_prime = 0,
  beta_A_L_prime = 0,
  beta_A_Z_prime = 0,
  beta_A_D_prime = 0,
  beta_A_C_prime = 0,
  beta_Z_L_prime = 0,
  beta_Z_A_prime = 0,
  beta_Z_D_prime = 0,
  beta_Z_C_prime = 0,
  beta_L0_L_prime = 0,
  beta_L0_A_prime = 0,
  beta_L0_Z_prime = 0,
  beta_L0_D_prime = 0,
  beta_L0_C_prime = 0,
  beta_A0_L_prime = 0,
  beta_A0_A_prime = 0,
  beta_A0_Z_prime = 0,
  beta_A0_D_prime = 0,
  beta_A0_C_prime = 0,
  ...
)

Arguments

N

Integer. Number of individuals to simulate.

eta

Numeric vector of length 4 (or 5). Shape parameters of the Weibull baseline intensity for each event type.

\eta \nu t^{\nu - 1}

.

nu

Numeric vector of length 4 (or 5). Scale parameters for the Weibull hazard.

adherence

Logical. Indicator of whether a Treatment process should be simulated.

followup

Numeric. Maximum censoring time. Events occurring after this time are censored. Default is Inf (no censoring).

cens

Logical. Indicator of whether there should be a censoring process.

generate.A0

Function. Function to generate the baseline treatment covariate A0. Takes N and L0 as inputs. Default is a Bernoulli(0.5) random variable.

lower

Numeric. Lower bound for root-finding in inverse cumulative hazard calculations. Default is 10^{-15}.

upper

Numeric. Upper bound for root-finding in inverse cumulative hazard calculations. Default is 200.

t_prime

Numeric scalar or NULL. Time point where effects change (optional).

at_risk_cov

Function. Function determining if an individual is at risk for each event type, given their covariates. Takes a numeric vector covariates and returns a binary vector. Default returns 1 for all events.

beta_L_A

Numeric. Specifies how L affects A.

beta_L_Z

Numeric. Specifies how L affects Z.

beta_L_D

Numeric. Specifies how L affects D.

beta_L_C

Numeric. Specifies how L affects C.

beta_A_L

Numeric. Specifies how L affects A.

beta_A_Z

Numeric. Specifies how L affects Z.

beta_A_D

Numeric. Specifies how L affects D.

beta_A_C

Numeric. Specifies how L affects C.

beta_Z_L

Numeric. Specifies how L affects A.

beta_Z_A

Numeric. Specifies how L affects Z.

beta_Z_D

Numeric. Specifies how L affects D.

beta_Z_C

Numeric. Specifies how L affects C.

beta_L0_L

Numeric. Specifies how L affects A.

beta_L0_A

Numeric. Specifies how L affects Z.

beta_L0_Z

Numeric. Specifies how L affects Z.

beta_L0_D

Numeric. Specifies how L affects D.

beta_L0_C

Numeric. Specifies how L affects C.

beta_A0_L

Numeric. Specifies how L affects A.

beta_A0_A

Numeric. Specifies how L affects Z.

beta_A0_Z

Numeric. Specifies how L affects Z.

beta_A0_D

Numeric. Specifies how L affects D.

beta_A0_C

Numeric. Specifies how L affects C.

beta_L_A_prime

Numeric. Specifies how L additionally affects A after time t_prime.

beta_L_Z_prime

Numeric. Specifies how L additionally affects Z after time t_prime.

beta_L_D_prime

Numeric. Specifies how L additionally affects D after time t_prime.

beta_L_C_prime

Numeric. Specifies how L additionally affects C after time t_prime.

beta_A_L_prime

Numeric. Specifies how L additionally affects A after time t_prime.

beta_A_Z_prime

Numeric. Specifies how L additionally affects Z after time t_prime.

beta_A_D_prime

Numeric. Specifies how L additionally affects D after time t_prime.

beta_A_C_prime

Numeric. Specifies how L additionally affects C after time t_prime.

beta_Z_L_prime

Numeric. Specifies how L additionally affects A after time t_prime.

beta_Z_A_prime

Numeric. Specifies how L additionally affects Z after time t_prime.

beta_Z_D_prime

Numeric. Specifies how L additionally affects D after time t_prime.

beta_Z_C_prime

Numeric. Specifies how L additionally affects C after time t_prime.

beta_L0_L_prime

Numeric. Specifies how L additionally affects after time A.

beta_L0_A_prime

Numeric. Specifies how L additionally affects after time Z.

beta_L0_Z_prime

Numeric. Specifies how L additionally affects after time Z.

beta_L0_D_prime

Numeric. Specifies how L additionally affects after time D.

beta_L0_C_prime

Numeric. Specifies how L additionally affects after time C.

beta_A0_L_prime

Numeric. Specifies how L additionally affects after time A.

beta_A0_A_prime

Numeric. Specifies how L additionally affects after time Z.

beta_A0_Z_prime

Numeric. Specifies how L additionally affects after time Z.

beta_A0_D_prime

Numeric. Specifies how L additionally affects after time D.

beta_A0_C_prime

Numeric. Specifies how L additionally affects after time C.

...

Additional arguments passed to simEventData or simEventTV

Details

simDropIn is a function that simulates data corresponding to N individuals that are at risk for 4 or 5 events. Censoring (C), Death (D), Drop In Initiation (Z), Change in Covariate Process (L) and optionally Treatment (A).

Value

Data frame containing the simulated event history data

Examples

simDropIn(10)

Simulate Event History Data Based on Cox Models

Description

Simulates recurrent and terminal event data for a cohort of individuals based on a list of fitted Cox proportional hazards models. Each event type is governed by its own model, and simulation proceeds by iteratively sampling event times until a terminal event occurs.

Usage

simEventCox(
  N,
  cox_fits,
  list_old_vars = NULL,
  n_event_max = c(1, 1),
  term_events = 1,
  intervention1 = NULL,
  intervention2 = NULL
)

Arguments

N

Integer. The number of individuals to simulate.

cox_fits

A named list of fitted Cox proportional hazards models (coxph objects), one for each event type. The names are used as event type labels.

list_old_vars

A named list containing the old covariates. New covariates will be simulated by drawing from the old covariates with replacement.

n_event_max

Integer vector. Maximum number of times each event type can occur per individual.

term_events

Integer or integer vector. Indices of event types that are terminal, i.e., events that stop further simulation for an individual.

intervention1

Optional function. Takes arguments ⁠(j, sim_matrix)⁠ and returns an updated simulation matrix. Used to modify covariates dynamically at each event iteration.

intervention2

Optional function. Takes arguments ⁠(j, H_j)⁠ and returns a modified baseline cumulative hazard vector for event type j. Allows dynamic hazard modification. The function intervention2 <- function(j, basehaz) if(j ==2) 1.15 * basehaz else basehaz performs an intervention where the baseline hazard of process 2 is multiplied by 1.15.

Details

The function simulates individual event histories by:

  1. Sampling initial baseline covariates by resampling observed values.

  2. Extracting baseline cumulative hazard functions from the Cox models.

  3. Iteratively sampling event times.

  4. Updating covariate histories and event counts.

  5. Stopping simulation per individual after a terminal event or maximum events reached.

Value

A data.table with one row per event per individual containing:

Examples

# The observed data
data_obs <- simDisease(N = 1000)
data_obs <- IntFormatData(data_obs, N_cols = 6)

# Fit Cox models
cox_death <- survival::coxph(survival::Surv(tstart, tstop, Delta == 1)
~ L0 + A0 + L, data = data_obs)
cox_Disease <- survival::coxph(survival::Surv(tstart, tstop, Delta == 2)
~ L0 + A0, data = data_obs[L == 0])

# Then simulate new data:
cox_fits <- list("D" = cox_death, "L" = cox_Disease)
list_old_vars <- list("L0" = data_obs$L0, "A0" = data_obs$A0)
new_data <- simEventCox(100, cox_fits = cox_fits, list_old_vars = list_old_vars)

Simulate Event Data with Multiple Event Types and Covariates

Description

Simulate Continuous Time-to-Event Data with Multiple Event Types

Usage

simEventData(
  N,
  beta = NULL,
  eta = NULL,
  nu = NULL,
  at_risk = NULL,
  term_deltas = c(0, 1),
  max_cens = Inf,
  add_cov = NULL,
  override_beta = NULL,
  max_events = 10,
  lower = 10^(-15),
  upper = 200,
  gen_A0 = NULL,
  gen_L0 = NULL,
  at_risk_cov = NULL,
  ...
)

Arguments

N

Integer. Number of individuals to simulate.

beta

Numeric matrix. Regression coefficients matrix where columns correspond to event types (N0, N1, ...) and rows correspond to covariates (L0, A0, L1, L2, ...) and event counts (N0, N1, ...). Default is a zero matrix.

eta

Numeric vector. Shape parameters of the Weibull baseline intensity for each event type. Default is 0.1 for all events.

nu

Numeric vector. Scale parameters of the Weibull baseline intensity for each event type. Default is 1.1 for all events.

at_risk

Function. Function determining if an individual is at risk for each event type, given their current event counts. Takes a numeric vector events and returns a binary vector. Default returns 1 for all events.

term_deltas

Integer vector. Event types considered terminal (after which no further events occur). Default is c(0, 1).

max_cens

Numeric. Maximum censoring time. Events occurring after this time are censored. Default is Inf (no maximal censoring).

add_cov

Named list of functions. Functions generating additional baseline covariates. Each function takes integer N and returns a numeric vector of length N. Default is NULL.

override_beta

Named list. Used to specify entries of the beta matrix to override defaults. For example, list("L0" = c("N1" = 2)) sets the effect of L0 on N1 to 2.

max_events

Integer. Maximum number of events to simulate per individual. Default is 10.

lower

Numeric. Lower bound for root-finding in inverse cumulative hazard calculations. Default is 10^{-15}.

upper

Numeric. Upper bound for root-finding in inverse cumulative hazard calculations. Default is 200.

gen_A0

Function. Function to generate the baseline treatment covariate A0. Takes N and L0 as inputs. Default is a Bernoulli(0.5) random variable.

gen_L0

Function. Function to generate the baseline covariate L0. Takes N as inputs. Default is a N(0,1) random variable.

at_risk_cov

Function. Function determining if an individual is at risk for each event type, given their covariates. Takes a numeric vector covariates and returns a binary vector. Default returns 1 for all events.

...

Additional technical arguments

Details

simEventData simulates event times and types for a cohort of individuals in a counting process framework. It supports multiple event types (by default 4), including terminal events, with intensities influenced by baseline covariates and previous event history.

The event intensities for event type x at time t are given by

\lambda^x(t) = \lambda_0^x(t) \exp(\beta_x^T L),

where the baseline intensity follows a Weibull hazard function:

\lambda_0^x(t) = \eta^x \nu^x t^{\nu^x - 1}.

Here, L is the vector of covariates and event counts, and \beta^x is the a vector of coefficients representing the effect of covariates and previous events on the intensity.

Value

A data.table with columns:

ID

Individual identifier

Time

Time of event

Delta

Event type at time

L0

Baseline covariate

A0

Baseline treatment

L1, L2, ...

Additional baseline covariates if specified

N0, N1, ...

Event counts up to the current event

Examples

# Simulate data for 10 individuals with default settings
sim_data <- simEventData(N = 10)
head(sim_data)

Simulate Survival and Competing Risk Data Based on a General Model

Description

The simEventObj function simulates survival or competing risk data for a cohort of individuals based on a general model with a predict2 method. The function is useful for simulating additional data under the same distribution as an original data set. The procedure consists of fitting a model, such as a random forest or Cox Proportional Hazards model on an original data set. Next the model is equipped with a predict2 method, and passed as an argument to the simEventObj function, which simulates new data using the predict2 method. The method should output the cumulative hazard array and the jump times of the cumulative hazard. Simulation proceeds by sampling from the uniform distribution and obtaining event times using the inverse of the cumulative hazard function(s).

Usage

simEventObj(N, obj, event_names = NULL, list_old_vars = NULL)

Arguments

N

Integer. The number of individuals to simulate.

obj

An object of class simevent. The object should have a predict2 method. The method should return a list containing chf and time. chf should be an array of dimension Individuals x Times x Events. The array should contain the cumulative hazard values for each individual, at each time for each event. time should be a vector of times where the cumulative hazard function jumps.

event_names

A character vector. Containing the names of the various processes. The argument is optional. By default events will be named N1, N2, ....

list_old_vars

A named list containing the old covariates. New covariates will be simulated by drawing from the old covariates with replacement.

Details

The function simulates individual event histories by:

  1. Sampling initial baseline covariates by resampling observed values.

  2. Extracting cumulative hazard functions from the object.

  3. Iteratively sampling event times.

  4. Updating covariate histories and event counts.

  5. Stopping simulation per individual after a terminal event or maximum events reached.

Value

A data.table with one row per event per individual containing:

simEventTV

Description

Simulate Event Data with Time-Varying Effects

Usage

simEventTV(
  N,
  beta = NULL,
  tv_eff = NULL,
  t_prime = Inf,
  eta = NULL,
  nu = NULL,
  at_risk = NULL,
  term_deltas = c(0, 1),
  max_cens = Inf,
  add_cov = NULL,
  override_beta = NULL,
  max_events = 10,
  lower = 10^(-15),
  upper = 200,
  gen_A0 = NULL,
  at_risk_cov = NULL
)

Arguments

N

Integer. Number of individuals to simulate.

beta

Matrix. Coefficients for covariates and processes. Columns correspond to events (N0, N1, ...), rows correspond to covariates (L0, A0, ..., and past event counts).

tv_eff

Matrix. Time-varying changes to beta, applied at time t_prime. Must have same dimensions as beta.

t_prime

Numeric. Time at which tv_eff is added to beta.

eta

Numeric vector. Shape parameters of Weibull intensities for each event.

nu

Numeric vector. Scale parameters of Weibull intensities for each event.

at_risk

Function. Determines which events an individual is at risk for, based on event history.

term_deltas

Integer vector. Event types considered terminal (e.g., death).

max_cens

Numeric. Maximum censoring time. Defaults to Inf.

add_cov

Named list of functions for generating additional baseline covariates. Each function takes one argument N and returns a vector of length N.

override_beta

Named list to override elements of beta. Format: list("covariate" = c("event" = value)).

max_events

Integer. Maximum number of events allowed per individual.

lower

Numeric. Lower bound for the root-finding algorithm used in inverse cumulative hazard computation.

upper

Numeric. Upper bound for the root-finding algorithm used in inverse cumulative hazard computation.

gen_A0

Function. Generates baseline treatment assignment. Takes arguments N and L0.

at_risk_cov

Function. Function determining if an individual is at risk for each event type, given their covariates. Takes a numeric vector covariates and returns a binary vector. Default returns 1 for all events.

Details

simEventTV is a function that simulates event data, with the option of adding time varying effects. The function is build up in the same way as simEventData, with the additional arguments tv_eff and t_prime, which specify the change of the beta matrix at time t_prime.

Value

A data.table with columns:

ID:

Individual identifier

Time:

Time of event

Delta:

Type of event

L0:

Baseline covariate

A0:

Baseline treatment

N0, N1, ...:

Cumulative event counts

L1, L2, ...:

Additional covariates (if specified)

Examples

eta <- rep(0.1, 2)
simEventTV(N = 100, t_prime = 1, eta = eta, term_deltas = c(0, 1))

Simulate Data in a Statin Setting

Description

Simulate Data in a Statin Setting

Usage

simStatinData(
  N,
  eta = rep(0.1, 6),
  nu = rep(1.1, 6),
  beta = NULL,
  cens = 1,
  followup = 5,
  lower = 10^(-15),
  upper = 200,
  gen_A0 = NULL,
  gen_L0 = NULL,
  ...
)

Arguments

N

Numeric scalar. Number of individuals to simulate.

eta

Numeric vector of length 6. Shape parameters for Weibull intensities with parameterization \eta \nu t^{\nu - 1}. Defaults to rep(0.1, 6).

nu

Numeric vector of length 6. Scale parameters for the Weibull hazards. Defaults to rep(1.1, 6).

beta

Numeric matrix. Of dimension p times 6. Regression coefficients matrix where columns correspond to event types (N0, ..., N5) and rows correspond to covariates (L0, A0, L1, L2, ...) followed by event counts (N0, ..., N5). Default is a zero matrix.

cens

Binary scalar. Indicates whether individuals are at risk of censoring (default 1).

followup

Numeric scalar. Maximum follow-up (censoring) time. Defaults to Inf.

lower

Numeric scalar. Lower bound for root-finding (inverse cumulative hazard) (default 1e-15).

upper

Numeric scalar. Upper bound for root-finding (default 200).

gen_A0

Function. Function to generate the baseline treatment covariate A0.Takes N and L0 as inputs. Default is a Bernoulli(0.5) random variable.

gen_L0

Function. Function to generate the baseline covariate L0. Takes N as input. Default is a N(0,1) random variable.

...

Additional arguments passed to simEventData

Value

A data frame containing the simulated data with columns:

ID

Individual identifier

Time

Time of the event

Delta

Event type (0,...,5)

L0

Baseline covariate

A0

Baseline covariate

L1, ...Lp

Additional baseline covariates

Examples

simDisease(10)

Simulate Survival Data

Description

Simulate Survival Data with Censoring and Event Times

Usage

simSurvData(N, beta = NULL, eta = rep(0.1, 2), nu = rep(1.1, 2), cens = 1, ...)

Arguments

N

Numeric scalar. Number of individuals to simulate.

beta

Numeric 2x2 matrix specifying effects of baseline covariates L0 and A0 on censoring and event hazards.

  • Rows correspond to covariates L0 and A0.

  • Columns correspond to censoring (1st column) and event (2nd column). Defaults to zero matrix if NULL.

eta

Numeric vector of length 2. Shape parameters for Weibull hazard with parameterization \eta \nu t^{\nu - 1}. Defaults to rep(0.1, 2).

nu

Numeric vector of length 2. Scale parameters for the Weibull hazard. Defaults to rep(1.1, 2).

cens

Numeric binary indicator (0 or 1) specifying if censoring is included (default 1).

...

Additional arguments passed to simEventData, including the argument add_cov to specify extra covariates.

Details

Simulates survival data for N individuals who are at risk for censoring (0) and an event (1). The hazard functions for censoring and event times follow Weibull distributions parameterized by shape parameters \eta and scale parameters \nu. Covariate effects on censoring and event hazards are specified via a matrix beta.

Value

Data frame containing the simulated survival data

Examples

simSurvData(10)

Simulate Event History Data with Treatment and Time-Dependent Covariate

Description

Simulates event history data with four types of events representing censoring (0), death (1), treatment (2), and covariate change (3). Death and censoring are terminal events; treatment and covariate events can occur only once.

Usage

simTreatment(
  N,
  eta = rep(0.1, 4),
  nu = rep(1.1, 4),
  beta_L_A = 1,
  beta_L_D = 1,
  beta_A_D = -1,
  beta_A_L = -0.5,
  beta_L0_A = 1,
  beta_L0_L = 1,
  beta_L0_D = 1,
  beta_L0_C = 0,
  beta_L_C = 0,
  beta_A_C = 0,
  beta_L_A_prime = 0,
  beta_L_D_prime = 0,
  beta_A_D_prime = 0,
  beta_A_L_prime = 0,
  beta_L0_A_prime = 0,
  beta_L0_L_prime = 0,
  beta_L0_D_prime = 0,
  beta_L0_C_prime = 0,
  beta_L_C_prime = 0,
  beta_A_C_prime = 0,
  t_prime = NULL,
  at_risk_cov = NULL,
  cens = 1,
  op = 1,
  lower = 10^(-15),
  upper = 200,
  followup = Inf,
  ...
)

Arguments

N

Integer. Number of individuals to simulate.

eta

Numeric vector of length 4. Shape parameters for Weibull intensities, parameterized as \eta \nu t^{\nu - 1}. Default is rep(0.1, 4).

nu

Numeric vector of length 4. Scale parameters for Weibull hazards. Default is rep(1.1, 4).

beta_L_A

Numeric. Effect of covariate L = 1 on treatment hazard. Default 1.

beta_L_D

Numeric. Effect of covariate L = 1 on death hazard. Default 1.

beta_A_D

Numeric. Effect of treatment A = 1 on death hazard. Default -1.

beta_A_L

Numeric. Effect of treatment A = 1 on covariate hazard. Default -0.5.

beta_L0_A

Numeric. Effect of baseline covariate L0 on treatment hazard. Default 1.

beta_L0_L

Numeric. Effect of baseline covariate L0 on covariate hazard. Default 1.

beta_L0_D

Numeric. Effect of baseline covariate L0 on death hazard. Default 1.

beta_L0_C

Numeric. Effect of baseline covariate L0 on censoring hazard. Default 0.

beta_L_C

Numeric. Effect of covariate L = 1 on censoring hazard. Default 0.

beta_A_C

Numeric. Effect of treatment A = 1 on censoring hazard. Default 0.

beta_L_A_prime

Numeric. Additional effect of covariate L = 1 on treatment hazard. Default 0.

beta_L_D_prime

Numeric. Additionalffect of covariate L = 1 on death hazard. Default 0.

beta_A_D_prime

Numeric. Effect of treatment A = 1 on death hazard. Default 0.

beta_A_L_prime

Numeric. Effect of treatment A = 1 on covariate hazard. Default 0.

beta_L0_A_prime

Numeric. Effect of baseline covariate L0 on treatment hazard. Default 0.

beta_L0_L_prime

Numeric. Effect of baseline covariate L0 on covariate hazard. Default 0.

beta_L0_D_prime

Numeric. Effect of baseline covariate L0 on death hazard. Default 0.

beta_L0_C_prime

Numeric. Effect of baseline covariate L0 on censoring hazard. Default 0.

beta_L_C_prime

Numeric. Effect of covariate L = 1 on censoring hazard. Default 0.

beta_A_C_prime

Numeric. Effect of treatment A = 1 on censoring hazard. Default 0.

t_prime

Numeric scalar or NULL. Time point where effects change (optional).

at_risk_cov

Function. Function determining if an individual is at risk for each event type, given their covariates. Takes a numeric vector covariates and returns a binary vector. Default returns 1 for all events.

cens

Integer (0 or 1). Indicates if censoring is possible. Default 1.

op

Integer (0 or 1). Indicates if treatment (operation) is possible. Default 1.

lower

Numeric. Lower bound for root finding (inverse cumulative hazard). Default 1e-15.

upper

Numeric. Upper bound for root finding (inverse cumulative hazard). Default 200.

followup

Numeric. Maximum censoring time. Defaults to Inf (no censoring).

...

Additional arguments passed to simEventData or simEventTV

Details

Event intensities are modeled using Weibull hazards with parameters \nu (scale) and \eta (shape), and covariate effects controlled by specified beta parameters. For example, beta_L_A quantifies the effect of covariate L = 1 on the hazard of treatment.

Value

A data.frame with columns:

Examples

simTreatment(10)