MCMC: MCMC sampling for count regression with a constant mean,...
In jayverhoef/POP: Platform of Opportunity Data and Models

Description Usage Arguments Value Author(s)

View source: R/MCMC.R

MCMC sampling for count regression with a constant mean, latent autocorrelated errors, and missing data

MCMC(
  response,
  datadist = dpois,
  antilink = exp,
  beta0ini,
  sigmaini = NULL,
  sigCARini,
  rhoini,
  zini,
  rhogrid,
  detsig,
  SigiList,
  eff_mat_offset = NULL,
  ind_samp,
  ind_miss,
  sample_sigma = FALSE,
  sample_sigCAR = TRUE,
  sample_rho = TRUE,
  use_trunc = FALSE,
  trunc_bound = NULL,
  nMCMC = 1000,
  n_z_iters = 10,
  n_thin = 1,
  samp_sd = 0.01,
  miss_sd = 0.01,
  beta0_sd = 0.1,
  sigma_sd = 0.1,
  sigCAR_sd = 0.02,
  rho_indxpm = 3
)

`response`	vector with response variable.
`datadist`	probability density function for data. It must be parameterized with a mean parameter, and a variance parameter. If there is no variance parameter, the argument must still be there, and just evaluated to NULL.
`antilink`	function to change mean (linear model) on link scale to mean for datadist distribution.
`beta0ini`	initialize beta0
`sigmaini`	initialize sigma for datadist, if it has one.
`sigCARini`	initialize sigCAR
`rhoini`	initialize rho. It must be one of the values in rhogrid
`zini`	initialize latent autocorrelated errors z
`rhogrid`	vector of rho values for lookup table
`detsig`	vector of determinant values corresponding to rhogrid
`SigiList`	list of inverse covariances, in sparse matrix form,
`eff_mat_offset`	matrix of samples from posterior distribution for effort. Default is NULL, in which case no offset is used
`ind_samp`	vector of indexes of sampled grid cells
`ind_miss`	vector of indexes of unsampled grid cells
`sample_sigma`	Logical. Should sigma parameter be sampled (TRUE) or held at initial values (FALSE).
`sample_sigCAR`	Logical. Should sigCAR parameter be sampled (TRUE) or held at initial values (FALSE).
`sample_rho`	Logical. Should rho parameter be sampled (TRUE) or held at initial values (FALSE).
`use_trunc`	Logical. Should truncation for z (latent autocorrelated errors) be used. If true, truncated to plus and minus the log of the largest observed value.
`nMCMC`	number of MCMC samples to retain
`n_z_iters`	number of latent autocorrelated errors (z) proposals per further MCMC updating
`n_thin`	number of MCMC samples per retained MCMC sample
`samp_sd`	tuning parameter for width of independent uniform proposal for Metropolis step of sampled z-values
`miss_sd`	tuning parameter for width of independent uniform proposal for Metropolis step of missing z-values
`beta0_sd`	tuning parameter for variance of independent normal proposal for Metropolis step of beta0
`sigma_sd`	tuning parameter for variance of independent normal proposal for Metropolis step of sigma
`sigCAR_sd`	tuning parameter for variance of independent normal proposal for Metropolis step of sigCAR
`rho_indxpm`	tuning parameter for number of adjacent grid values, plus and minus, of proposal for Metropolis/Hastings step of rho