agfh (version 0.2.1)

make_agfh_sampler: Maker Function: Agnostic Fay-Herriot Sampler

Description

A maker function that returns a function. The returned function is a sampler for the agnostic Fay-Herriot model.

Value

Returns a sampler, itself a function of initial parameter values (a list with values for \(\beta, \theta\), the latent variance of \(\theta\), and starting values for \(g(.)\), typically zeros), number of samples, thinning rate, and scale of Metropolis-Hastings jumps for \(\theta\) sampling.

Arguments

X

observed independent data to be analyzed

Y

observed dependent data to be analyzed

D

known precisions of response Y

var_gamma_a

latent variance prior parameter, rgamma shape

var_gamma_b

latent variance prior parameter, rgamma rate

S

vector of starting support values for \(g(\cdot)\)

kern.a0

scalar variance parameter of GP kernel

kern.a1

scalar lengthscale parameter of GP kernel

kern.fuzz

scalar noise variance of kernel

Details

Creates a Metropolis-within-Gibbs sampler of the agnostic Fay-Herriot model (AGFH).

Examples

  n <- 10
  X <- matrix(1:n, ncol=1)
  Y <- 2*X + rnorm(n, sd=1.1)
  D <- rep(1, n)
  ag <- make_agfh_sampler(X, Y, D)

  params.init <- list(
    beta=1,
    theta=rep(0,n),
    theta.var=1,
    gamma=rep(0,n)
  )
  ag.out <- ag(params.init, 5, 1, 0.1)