Simulate count / relative-abundance GLMM data — simulate

One observed value per site from a GLMM with no detection process, matching the count() family: log-link Poisson / negative-binomial counts, an identity-link Gaussian response, or a logit-link binomial response (k successes out of n trials per site). The mean predictor is X beta with beta the fixed-effect coefficients (intercept first).

Usage

simulate_count(
  N = 200,
  beta = c(1, 0.5),
  response = c("poisson", "negbin", "gaussian", "binomial"),
  size = 2,
  sd = 1,
  trials = 10,
  seed = NULL
)

Arguments

N: Number of sites (default 200).
beta: Fixed-effect coefficients, length 1 + n_covs (intercept first).
response: The response distribution: "poisson", "negbin", "gaussian", or "binomial".
size: Negative-binomial size (dispersion); larger is closer to Poisson. Used only for response = "negbin".
sd: Gaussian residual SD. Used only for response = "gaussian".
trials: Per-site trial count for response = "binomial": a scalar (recycled) or a length-N vector. Default 10 (a scalar 1 gives Bernoulli data, the svcPGBinom trials = 1 setting).
seed: Random seed.

Value

A list with y (a length-N numeric response vector; success counts for the binomial response), data (the site covariates), and truth (the coefficients, response, dispersion, and the binomial trials).

Examples

sim <- simulate_count(N = 100, beta = c(1, 0.5), seed = 1)
length(sim$y)