Simulate Royle (2004) N-mixture abundance data — simulate

Latent abundance N_i ~ Poisson(lambda_i) (or, under mixture = "negbin", N_i ~ NegBin(mean = lambda_i, size = size) with variance lambda + lambda^2 / size) with log lambda_i = X_lambda beta_lambda, and replicate counts y_ij ~ Binomial(N_i, p_i) with logit p_i = X_p beta_p (site-level detection). The returned y is an N x J integer count matrix suitable for tobs() with abun().

Usage

simulate_abun(
  N = 100,
  J = 4,
  n_abund_covs = 2,
  n_det_covs = 1,
  beta_lambda = NULL,
  beta_p = NULL,
  mixture = c("poisson", "negbin", "zip", "zinb"),
  size = 2,
  omega = 0.3,
  seed = NULL
)

Arguments

N: Number of sites (default 100).
J: Number of replicate visits (default 4).
n_abund_covs: Number of abundance covariates (default 2).
n_det_covs: Number of detection covariates (default 1).
beta_lambda: Abundance coefficients c(intercept, slopes...) on the log scale. Default c(log(3), runif(n_abund_covs, -0.5, 0.5)).
beta_p: Detection coefficients c(intercept, slopes...) on the logit scale. Default c(0, runif(n_det_covs, -0.5, 0.5)) (intercept 0 = p 0.5).
mixture: Abundance mixing distribution: "poisson" (default), "negbin" (negative binomial, overdispersed), or their zero-inflated counterparts "zip" / "zinb" (a structural-zero share omega of sites have N = 0 regardless of lambda).
size: Negative-binomial size r (the "negbin" / "zinb" dispersion; variance lambda + lambda^2 / r). Smaller r means more overdispersion; as r grows large the model approaches Poisson. Default 2. Ignored under Poisson.
omega: Structural-zero probability for "zip" / "zinb" (the share of sites with N = 0 independent of lambda). Default 0.3. Ignored otherwise.
seed: Optional random seed.

Value

A list with y (N x J count matrix), data (covariate data frame), and truth (the coefficients, per-site lambda, p, latent N, and the mixture / size used).

Examples

sim <- simulate_abun(N = 50, J = 3, seed = 1)
dim(sim$y)