LangevinGibbs: Langevin-Gibbs sampler for inference in a mixture cure model.

In oswaldogressani/mixcurelps: Approximate Bayesian Inference in Mixture Cure Models

Langevin-Gibbs sampler for inference in a mixture cure model.

Description

This routine implements a Metropolis-Langevin-within-Gibbs sampler to draw samples from the posterior distribution of a mixture cure model. A Metropolis-adjusted Langevin algorithm is used to sample from the conditional posterior distribution of the latent vector. MCMC samples for the roughness penalty parameter and the dispersion parameter are obtained via a Gibbs sampler.

Usage

LangevinGibbs(
  formula,
  data,
  K = 15,
  penorder = 3,
  deltaprior = 1e-04,
  mcmcsample = 10000,
  burnin = 2000,
  tunparam = 0.25,
  mcmcseed = NULL,
  progbar = c("yes", "no")
)

Arguments

formula	A model formula of the form Surv(tobs,delta)~ inci()+late().
data	A data frame.
K	The number of B-spline coefficients.
penorder	The order of the penalty associated to the B-spline coefficients.
deltaprior	The parameters of the Gamma prior for the dispersion parameter.
mcmcsample	The length of the MCMC chain.
burnin	The length of the burnin.
tunparam	The initial tuning parameter of the MLWG algorithm, default is 0.25.
mcmcseed	The seed to be used (for reproducibility).
progbar	Should a progress bar be shown? Default is yes.

Author(s)

Oswaldo Gressani [e-mail] .

oswaldogressani/mixcurelps documentation built on Oct. 30, 2024, 10:45 p.m.