Fits the specified longitudinal model for time-varying covariates up to the landmark times

Usage

fit_longitudinal(
  x,
  landmarks,
  method,
  formula,
  dynamic_covariates,
  validation_fold = 0,
  cores = getOption("Ncpus", 1L),
  .warn_when_prop_few_obs = 0.25,
  ...
)

Arguments

x: An object of class LandmarkAnalysis.
landmarks: A vector of Landmark times.
method: Either "lcmm" or "lme4" or a function for fitting a longitudinal data model, where the first argument is a formula, and also has a data argument. Only needed for fit-based prediction methods used later in predict_longitudinal (e.g. "lcmm", "lme4"); summary measures such as "locf" are computed directly from the data and do not require a call to fit_longitudinal.
formula: A formula to be used in longitudinal sub-model fitting.
dynamic_covariates: Vector of time-varying covariates to be modelled as the outcome of a longitudinal model.
validation_fold: If positive, cross-validation fold where model is fitted. If 0 (default), model fitting is performed using the complete dataset.
cores: Number of cores/threads to be used for parallel computation on Linux and MacOS. Defaults to either options("Ncpus") if set, or 1 (single threaded) otherwise. Only single-threaded computation is currently supported on Windows.
.warn_when_prop_few_obs: Threshold proportion (0-1) for warning when individuals have 0 or 1 observations. Defaults to 0.25 (i.e., warn when 25% or more individuals have few observations).
...: Additional arguments passed to the longitudinal model fitting function (e.g. number of classes/clusters for lcmm).

Value

An object of class LandmarkAnalysis.

Details

Parallel processing

As the longitudinal model for each landmark time is independent of the longitudinal models for other landmark times, parallel processing can be used to vastly speed up computation. Currently only Unix-like operating systems are supported by landmaRk.