Low-level Functions to generate draws from the posterior distribution of model coefficients

Generate posterior draws from a fitted model

Usage

post_draws(model, ...)

# S3 method for default
post_draws(
  model,
  n,
  method = c("gaussian", "mh", "inla", "user"),
  mu = NULL,
  sigma = NULL,
  n_cores = 1L,
  burnin = 1000,
  thin = 1,
  t_df = 40,
  rw_scale = 0.25,
  index = NULL,
  frequentist = FALSE,
  unconditional = FALSE,
  parametrized = TRUE,
  mvn_method = c("mvnfast", "mgcv"),
  draws = NULL,
  seed = NULL,
  ...
)

generate_draws(model, ...)

# S3 method for gam
generate_draws(
  model,
  n,
  method = c("gaussian", "mh", "inla"),
  mu = NULL,
  sigma = NULL,
  n_cores = 1L,
  burnin = 1000,
  thin = 1,
  t_df = 40,
  rw_scale = 0.25,
  index = NULL,
  frequentist = FALSE,
  unconditional = FALSE,
  mvn_method = c("mvnfast", "mgcv"),
  seed = NULL,
  ...
)

Arguments

model: a fitted R model. Currently only models fitted by mgcv::gam() or mgcv::bam(), or return an object that inherits from such objects are supported. Here, "inherits" is used in a loose fashion; models fitted by scam::scam() are support even though those models don't strictly inherit from class "gam" as far as inherits() is concerned.
...: arguments passed to methods.
n: numeric; the number of posterior draws to take.
method: character; which algorithm to use to sample from the posterior. Currently implemented methods are: "gaussian" and "mh". "gaussian" calls gaussian_draws() which uses a Gaussian approximation to the posterior distribution. "mh" uses a simple Metropolis Hasting sampler which alternates static proposals based on a Gaussian approximation to the posterior, with random walk proposals. Note, setting t_df to a low value will result in heavier-tailed statistic proposals. See mgcv::gam.mh() for more details.
mu: numeric; user-supplied mean vector (vector of model coefficients). Currently ignored.
sigma: matrix; user-supplied covariance matrix for mu. Currently ignored.
n_cores: integer; number of CPU cores to use when generating multivariate normal distributed random values. Only used if mvn_method = "mvnfast" and method = "gaussian".
burnin: numeric; the length of any initial burn in period to discard. See mgcv::gam.mh().
thin: numeric; retain only thin samples. See mgcv::gam.mh().
t_df: numeric; degrees of freedom for static multivariate t proposal. See mgcv::gam.mh().
rw_scale: numeric; factor by which to scale posterior covariance matrix when generating random walk proposals. See mgcv::gam.mh().
index: numeric; vector of indices of coefficients to use. Can be used to subset the mean vector and covariance matrix extracted from model.
frequentist: logical; if TRUE, the frequentist covariance matrix of the parameter estimates is used. If FALSE, the Bayesian posterior covariance matrix of the parameters is used. See mgcv::vcov.gam().
unconditional: logical; if TRUE the Bayesian smoothing parameter uncertainty corrected covariance matrix is used, if available for model. See mgcv::vcov.gam().
parametrized: logical; use parametrized coefficients and covariance matrix, which respect the linear inequality constraints of the model. Only for scam::scam() model fits.
mvn_method: character; one of "mvnfast" or "mgcv". The default is uses mvnfast::rmvn(), which can be considerably faster at generate large numbers of MVN random values than mgcv::rmvn(), but which might not work for some marginal fits, such as those where the covariance matrix is close to singular.
draws: matrix; user supplied posterior draws to be used when method = "user".
seed: numeric; the random seed to use. If NULL, a random seed will be generated without affecting the current state of R's RNG.