Fit Centroid MVN Small Area Estimation model using tmbstan.

Random effects have a multivariate Gaussian distribution with covariance matrix calculated using centroid_covariance. Kernel hyper-parameters are fixed.

Usage

fck_tmbstan(
  sf,
  nsim_warm = 100,
  nsim_iter = 1000,
  chains = 4,
  cores = parallel::detectCores(),
  kernel = matern,
  ii = NULL,
  ...
)

Arguments

sf: A simple features object with some geometry.
nsim_warm: Number of warmup samples, passed to rstan.
nsim_iter: Number of samples, passed to rstan.
chains: Number of chains, each of which gets nsim_warm + nsim_iter samples, passed to rstan.
cores: Number of cores, passed to rstan, defaults to parallel::detectCores().
kernel: A kernel function, defaults to matern.
ii: The (zero-indexed) indices of the observations held-out.
...: Additional arguments to kernel.

Examples

fck_tmbstan(mw, nsim_warm = 0, nsim_iter = 100, cores = 2)
#> Error in sf::st_centroid(sf): object 'mw' not found