Fit a hierarchical biomarker kinetics model — fit

Fits the epikinetics piecewise-linear model with CmdStanR and returns a self-contained S3 fit object. The likelihood supports uncensored, left- censored, and right-censored observations. Parallel chains and Stan threads within each chain can be used together. Data validation and transformation are deliberately performed beforehand by prepare_epikinetics_data().

Usage

fit_epikinetics(
  model_data,
  chains = 4,
  parallel_chains = chains,
  threads_per_chain = 1,
  grainsize = NULL,
  adapt_delta = 0.9,
  max_treedepth = 12,
  ...
)

Arguments

model_data: An object returned by prepare_epikinetics_data().
chains: Number of Markov chains.
parallel_chains: Number of chains to run concurrently.
threads_per_chain: Number of Stan threads used by each chain. The model is always compiled with threading enabled.
grainsize: Number of participants handled by each reduce_sum task. By default it is chosen from the participant and thread counts.
adapt_delta: Target acceptance probability. The conservative default is appropriate for this nonlinear hierarchy and may be overridden after inspecting diagnostics.
max_treedepth: Maximum NUTS tree depth.
...: Additional arguments passed to the CmdStanR model's $sample() method, such as iter_warmup, iter_sampling, and seed.

Value

An object of class epikinetics_fit.

Examples

if (FALSE) { # \dontrun{
dat <- utils::read.csv(
  system.file("extdata", "delta.csv", package = "epikinetics")
)
model_data <- prepare_epikinetics_data(
  dat,
  formula = ~ infection_history,
  lower_limit = 5,
  upper_limit = 2560
)
model_data
model.matrix(model_data)
stan_data(model_data)

fit <- fit_epikinetics(
  model_data,
  chains = 4,
  parallel_chains = 4,
  threads_per_chain = 2
)
} # }