simMultiJM {MJMbamlss}

New Simulation Function For Multivariate JMs Based On FPCs

Description

Adapt the structure given by simJM function in bamlss.

Usage

simMultiJM(
  nsub = 300,
  times = seq(0, 120, 1),
  probmiss = 0.75,
  max_obs = length(times),
  maxfac = 1.5,
  nmark = 2,
  long_assoc = c("FPC", "splines", "param"),
  M = 6,
  FPC_bases = NULL,
  FPC_evals = NULL,
  mfpc_args = list(type = "split", eFunType = "Poly", ignoreDeg = NULL, eValType =
    "linear", eValScale = 1),
  re_cov_mat = NULL,
  ncovar = 2,
  lambda = function(t, x) {
     1.4 * log((t + 10)/1000)
 },
  gamma = function(x) {
     -1.5 + 0.3 * x[, 1]
 },
  alpha = rep(list(function(t, x) {
     0.3 + 0 * t
 }), nmark),
  mu = rep(list(function(t, x) {
     1.25 + 0.6 * sin(x[, 2]) + (-0.01) * t
 }), nmark),
  sigma = function(t, x) {
     0.3 + 0 * t + I(x$marker == "m2") * 0.2
 },
  tmax = NULL,
  seed = NULL,
  full = FALSE,
  file = NULL
)

Arguments

nsub

Number of subjects.

times

Vector of time points.

probmiss

Probability of missingness.

max_obs

Maximal number of observations per individual and marker. Defaults to no upper limit.

maxfac

Factor changing the uniform censoring interval.

nmark

Number of markers.

long_assoc

Longitudinal association between the markers (Defaults to "FPC"). If "splines" or "param", then specify the normal covariance matrix with argument 're_cov_mat' and include the random effects in argument mu. If "FPC", then principal components are used to model the association structure.

M

Number of principal components.

FPC_bases

FunData object. If supplied, use the contained FPC as basis for the association structure.

FPC_evals

Vector of eigenvalues. If supplied, use the provided eigenvalues for the association structure.

mfpc_args

List containing the named arguments "type", "eFunType", "ignoreDeg", "eValType" of function simMultiFunData and "eValScale" for scaling the eigenvalues.

re_cov_mat

If supplied, a covariance matrix to use for drawing the random effects needed for the association structure.

ncovar

Number of covariates.

lambda

Additive predictor of time-varying survival covariates.

gamma

Additive predictor of time-constant survival covariates.

alpha

List of length nmark containing the additive predictors of the association.

mu

List of length nmark containing the additive predictors of the longitudinal part.

sigma

Additive predictor of the variance.

tmax

Maximal time point of observations.

seed

Seed for reproducibility.

full

Create a wide-format data.frame and a short one containing only survival info.

file

Name of the data file the generated data set should be stored into (e.g., "simdata.RData") or NULL if the dataset should directly be returned in R.

Value

For full = TRUE a list of four data.frames is returned:

data

Simulated dataset in long format including all longitudinal and survival covariates.

data_full

Simulated dataset on a grid of fixed time points.

data_hypo

Simulated dataset on a grid of fixed time points with hypothetical longitudinal outcomes after the event.

fpc_base

If applicable, include the FPC basis used for simulation.

data_short

Convenience output containing only one observation per subject for easy access to event-times.

For full = FALSE only the first dataset is returned.

Examples

# Number of individuals
n <- 15
# Covariance matrix for the data generation
auto <- matrix(c(0.08, -0.07, -0.07, 0.9), ncol = 2)
cross <- matrix(rep(0.03, 4), ncol = 2)
cor <- matrix(c(0, 1, 0.75, 0.5, 0, 0,
                1, 0, 1, 0.75, 0.5, 0,
                0.75, 1, 0, 1, 0.75, 0.5,
                0.5, 0.75, 1, 0, 1, 0.75,
                0, 0.5, 0.75, 1, 0, 1,
                0, 0, 0.5, 0.75, 1, 0),
              ncol = 6)
cov <- kronecker(cor, cross) +
    kronecker(diag(c(1, 1.2, 1.4, 1.6, 1.8, 2)), auto)

# Simulate the data
d_rirs <- simMultiJM(
  nsub = n, times = seq(0, 1, by = 0.01), max_obs = 15, probmiss = 0.75,
  maxfac = 1.75, nmark = 6, long_assoc = "param", M = NULL, FPC_bases = NULL,
  FPC_evals = NULL, mfpc_args = NULL, re_cov_mat = cov, ncovar = 2,
  lambda = function(t, x) {1.37 * t^(0.37)},
  gamma = function(x) {-1.5 + 0.48*x[, 3]},
  alpha = list(function(t, x) {1.5 + 0*t}, function(t, x) {0.6 + 0*t},
               function(t, x) {0.3 + 0*t}, function(t, x) {-0.3 + 0*t},
               function(t, x) {-0.6 + 0*t}, function(t, x) {-1.5 + 0*t}),
  mu = list(function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 1] + r[, 2]*t
  }, function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 3] + r[, 4]*t
  }, function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 5] + r[, 6]*t
  }, function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 7] + r[, 8]*t
  }, function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 9] + r[, 10]*t
  }, function(t, x, r){
    0 + 0.2*t - 0.25*x[, 3] - 0.05*t*x[, 3] + r[, 11] + r[, 12]*t
  }),
  sigma = function(t, x) {log(0.06) + 0*t}, tmax = NULL, seed = NULL,
  full = TRUE, file = NULL)