simulate_tvem_example: simulate_tvem_example: Simulate a dataset for demonstrating the tvem function.

Description

By default, the data-generating model has a time-varying intercept, and two time-varying covariates named x1 and x2. x1 has a time-varying effect and x2 has a time-invariant effect.

Usage

simulate_tvem_example(
  n_subjects = 300,
  max_time = 7,
  simulate_binary = FALSE,
  n_obs_possible = 141,
  prop_obs_observed = 0.3,
  sigma_x1 = 2,
  sigma_x2 = 2,
  truncate_for_realism = TRUE,
  round_digits = 3,
  x1_short_term_rho = 0.7,
  x2_short_term_rho = 0.7,
  sigma_y = 1.5,
  mu_x1_function = function(t) {
     6 - 2 * sqrt(pmax(0, (t/7) - 0.2))
 },
  mu_x2_function = function(t) {
     3 + sqrt(pmax(0, (t/7) - 0.5))
 },
  beta0_y_function = function(t) {
     1 - 0.3 * sqrt(t/7)
 },
  beta1_y_function = function(t) {
     0.5 * (t/7)^2
 },
  beta2_y_function = function(t) {
     rep(0.2, length(t))
 },
  y_short_term_rho = 0.8
)

Value

A simulated dataset with the following variables:

subject_id: Subject ID
time: Observation time
x1: First covariate
x2: Second covariate
y: Outcome variable

Arguments

n_subjects: Number of subjects in dataset
max_time: The time point at the end of the simulated time interval
simulate_binary: Whether the simulated data should be binary
n_obs_possible: Total number of possible measurement times per subject
prop_obs_observed: Proportion of these that are actually observed
sigma_x1: Standard deviation of covariate 1, assumed homoskedastic over time
sigma_x2: Standard deviation of covariate 2, assumed homoskedastic over time
truncate_for_realism: Whether to prevent simulated values from going below 0 or above 10, in order to imitate survey data; used only for normally distributed outcomes. Set this to FALSE if you are running a simulation, in order to avoid losing coverage due to departing from the parametric model.
round_digits: Number of digits at which to round the generated data; used only for normally distributed outcomes
x1_short_term_rho: Correlation between adjacent measurements of covariate 1
x2_short_term_rho: Correlation between adjacent measurements of covariate 2
sigma_y: Error standard deviation of y, only used if the outcomes are to be normal rather than binary
mu_x1_function: Mean of covariate 1 as function of time
mu_x2_function: Mean of covariate 2 as function of time
beta0_y_function: TVEM intercept as function of time
beta1_y_function: TVEM coefficient of covariate 1 as function of time
beta2_y_function: TVEM coefficient of covariate 2 as function of time
y_short_term_rho: Correlation between adjacent measurements of y, only used if it is normal and not binary

Examples

Run this code

set.seed(123)
the_data <- simulate_tvem_example(simulate_binary=TRUE)

Run the code above in your browser using DataLab