Set priors for the occARU model

Constructs and validates a named list of prior hyperparameters for use with fit_model(). Any unspecified priors use the defaults listed below.

Usage

set_priors(
  psi_bar = c(1, 1),
  mu_bar = c(1, 1),
  psi_W = c(3, 0, 1),
  mu_W = c(3, 0, 2.5),
  psi_theta = c(1, 1),
  mu_theta = c(1, 1),
  iota_ell = c(1, 1),
  kappa_ell = c(1, 1),
  kappa_ell_periodic = c(1, 1),
  K_phi = c(1, 1),
  phi = c(0.4, 0.3),
  alpha_O_L = 1,
  psi_beta_O_L = 1,
  mu_beta_O_L = 1,
  gamma_O_L = 1,
  iota_O_L = 1,
  kappa_O_L = 1,
  epsilon_O_L = 1,
  verbose = TRUE
)

Arguments

psi_bar: Numeric vector of length 2. c(a, b) for a Beta(a, b) prior on mean occupancy probability. Default: c(1, 1).
mu_bar: Numeric vector of length 2. c(shape, rate) for a Gamma(shape, rate) prior on mean detection rate. Default: c(1, 1).
psi_W: Numeric vector of length 3. c(df, mu, sigma) for a Student-t+(df, mu, sigma) prior on total occupancy log odds variance. Default: c(3, 0, 1).
mu_W: Numeric vector of length 3. c(df, mu, sigma) for a Student-t+(df, mu, sigma) prior on total log detection variance. Default: c(3, 0, 2.5).
psi_theta: Numeric vector of length 2. c(shape, rate) for a Gamma(shape, rate) prior on the occupancy variance partition sparsity parameter. Default: c(1, 1).
mu_theta: Numeric vector of length 2. c(shape, rate) for a Gamma(shape, rate) prior on the detection variance partition sparsity parameter. Default: c(1, 1).
iota_ell: Numeric vector of length 2. c(alpha, beta) for an InvGamma(alpha, beta) prior on the spatial GP length scale(s). Only used when spatial = "gp" in fit_model(). Default: c(1, 1).
kappa_ell: Numeric vector of length 2. c(alpha, beta) for an InvGamma(alpha, beta) prior on the exp. quad. temporal GP length scale. Only used when temporal = "gp" in fit_model(). Default: c(1, 1).
kappa_ell_periodic: Numeric vector of length 2. c(alpha, beta) for an InvGamma(alpha, beta) prior on the periodic temporal GP length scale. Only used when temporal = "gp" and periodic = TRUE in fit_model(). Default: c(1, 1).
K_phi: Numeric vector of length 2. c(alpha[1], alpha[2]) for a Dirichlet(alpha) prior on the temporal GP variance partitions of the exp. quad. and periodic kernels. Only used when temporal_gp = TRUE and periodic_GP = TRUE. Default: c(1, 1).
phi: Numeric vector of length 2. c(alpha, beta) for an InvGamma(alpha, beta) prior on species-specific negative binomial overdispersion parameters. Only used when overdispersion = "nb" in fit_model(). Default: c(0.4, 0.3).
alpha_O_L: Positive scalar. LKJ prior on the \([2, 2]\) occupancy log odds/log detection rate correlation matrix. Default: 1.
psi_beta_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific occupancy site coefficients. Only used when site predictors are supplied for occupancy in make_data(). Default: 1.
mu_beta_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific detection site coefficients. Only used when site predictors are supplied for detection in make_data(). Default: 1.
gamma_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific survey coefficients. Only used when survey predictors are supplied. Default: 1.
iota_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific site effects. Default: 1.
kappa_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific survey effects. Default: 1.
epsilon_O_L: Positive scalar. LKJ prior on the \([S, S]\) correlation matrix of species-specific OLRE residuals. Only used when overdispersion = "olre" in fit_model(). Default: 1.
verbose: Logical. If TRUE (default), prints list of priors.

Value

An occARU_priors object (a named list) for use with fit_model().

Usage

Arguments

Value

See also