Generate synthetic longitudinal data with optional right-censoring

This function simulates repeated measurements of normally-distributed covariates, binary treatments, and an end-of-study outcome for longitudinal causal analyses. When `right_censor = TRUE`, a right-censoring indicator `Cj` is generated at each visit: if `Cj = 1`, all subsequent `L`, `A`, and `Y` values are set to `NA`.

Usage

simData(
  n,
  n_visits,
  covariate_counts = rep(2, n_visits),
  amodel,
  ymodel,
  y_type = c("binary", "continuous"),
  right_censor = FALSE,
  cmodel = NULL,
  seed = NULL
)

Arguments

n

Integer. Sample size.

n_visits

Integer. Number of visits (including baseline as visit 1).

covariate_counts

Integer vector of length `n_visits`. Number of covariates per visit (default: rep(2, n_visits)).

amodel

List of length `n_visits`. Each element is a named numeric vector of coefficients for the logistic model of treatment `Aj` on covariates (and `A_prev` for j > 1).

ymodel

Named numeric vector. Coefficients for the end-of-study outcome model. If `y_type = "binary"`, a logistic model is used; if `"continuous"`, a linear model with Gaussian noise.

y_type

Character. One of "binary" or "continuous".

right_censor

Logical. If TRUE, generates `Cj` using `cmodel` at each visit.

cmodel

List of length `n_visits`. Named numeric vectors for logistic censoring models at each visit, regressing `Cj` on covariates and current `Aj`.

seed

Integer. Optional random seed.

Value

A `data.frame` with columns `Lk_j`, `Aj`, optional `Cj`, and `Y`.