Simulate community (multispecies) joint occupancy-cover data

Per-species joint occupancy-detection + cover model with Gaussian community hyperpriors on the per-species coefficients of all three arms: beta_occ_s ~ N(mu_occ, diag(sd_occ^2)), beta_p_s ~ N(mu_p, diag(sd_p^2)), beta_pos_s ~ N(mu_pos, diag(sd_pos^2)), then per cell z_{s,i} ~ Bernoulli(psi_{s,i}), per visit y_{s,i,j} | z = 1 ~ Bernoulli(p_{s,i,j}), and per detected visit c_{s,i,j} ~ f_pos(eta_pos_{s,i,j}, disp). The site and visit covariates are shared across species (community covariates). The returned y / y_pos are 3D arrays [n_sites x J x n_species] suitable for tobs() with ms_occu_cover() (y_pos is NA where not detected).

Usage

simulate_ms_occu_cover(
  n_species = 12,
  N = 120,
  J = 5,
  n_occ_covs = 1,
  n_det_covs = 1,
  n_pos_covs = 1,
  mu_occ = NULL,
  mu_p = NULL,
  mu_pos = NULL,
  sd_occ = 0.5,
  sd_p = 0.4,
  sd_pos = 0.4,
  positive = c("lognormal", "beta", "gaussian"),
  phi = 30,
  sigma_pos = 0.4,
  seed = NULL
)

Arguments

n_species

Number of species (default 12).

N

Number of sites / cells (default 120).

J

Number of replicate visits (default 5).

n_occ_covs, n_det_covs, n_pos_covs

Number of covariates on each arm (drawn IID standard normal, shared across species).

mu_occ, mu_p, mu_pos

Community-mean coefficient vectors c(intercept, slopes...) on each arm's link scale. Defaults pick weakly-informative values.

sd_occ, sd_p, sd_pos

Per-coefficient community SD on each arm (length 1, recycled, or one per coefficient). Default 0.5 / 0.4 / 0.4.

positive

"lognormal" (default) or "beta".

phi

Beta precision when positive = "beta" (default 30).

sigma_pos

Lognormal residual SD when positive = "lognormal" (default 0.4).

seed

Optional integer seed.

Value

A list with y (3D detection array), y_pos (3D cover array, NA where not detected), data (per-cell covariate frame), visit_data (per-visit covariate frame, N*J rows in site-major order), species (species names), and truth (community means / SDs, per-species coefficients, the dispersion, and the latent state).

Examples

sim <- simulate_ms_occu_cover(n_species = 4, N = 30, J = 3, seed = 1)
dim(sim$y)