Simulate Hierarchical Negative Binomial Data — sim_hier

Generates simulated data from a hierarchical negative binomial model with optional observed heterogeneity via Z covariates.

Usage

sim_hier_negbin(
  nreg = 100,
  nobs = 50,
  nvar = 2,
  nz = 3,
  const = TRUE,
  het_observed = c("none", "linear", "step"),
  target_var_betabar = 0.5,
  target_var_eps = 0.25,
  alpha = 5,
  seed = NULL
)

Arguments

nreg: Number of cross-sectional units.
nobs: Number of observations per unit.
nvar: Number of X variables (excluding intercept if const=TRUE).
nz: Number of Z variables. Set to 0 for no observed heterogeneity.
const: Logical. Include an intercept in X? Default TRUE.
het_observed: Character. Functional form of observed heterogeneity. Options: "none", "linear", "step".
target_var_betabar: Numeric. Target variance for the first coefficient of the observed component. Default 0.5.
target_var_eps: Numeric. Target variance for each coefficient in the unobserved component eps_i. Default 0.25.
alpha: Numeric. Overdispersion parameter for negative binomial. Default 5.
seed: Integer. Optional random seed.

Value

A list containing:

regdata: List of length nreg. Each element has y, X.
hessdata: List of length nreg. Each element has hess (approx Hessian).
Z: nreg x nz matrix of covariates (NULL if nz=0).
true_values: List with beta_true, betabar_true, alpha, etc.