estimate_zinarp {ZINARp} | R Documentation |
Parameter estimation for ZINARp models
Description
This function uses MCMC algorithms (Metropolis-Hastings and Gibbs Sampler) to generate a chain of INAR/ZINAR(p) parameter estimators.
Usage
estimate_zinarp(
x,
p,
iter = 5000,
thin = 2,
burn = 0.1,
innovation = "Poisson"
)
Arguments
x |
A vector containing a discrete non-negative time series dataset. |
p |
The order of the INAR/ZINAR process. |
iter |
The number of iterations to be considered. Defaults to 5000. |
thin |
Lag for posterior sample. Defaults to 2. |
burn |
Burn-in for posterior sample. Defaults to 0.1. Must be in (0,1). |
innovation |
Distribution to be used for the innovation : "Poisson" or "ZIP". Defaults to Poisson. |
Value
Returns a list containing a posteriori samples for the specified model parameters.
References
Garay, Aldo M., Francyelle L. Medina, Celso RB Cabral, and Tsung-I. Lin. "Bayesian analysis of the p-order integer-valued AR process with zero-inflated Poisson innovations." Journal of Statistical Computation and Simulation 90, no. 11 (2020): 1943-1964.
Garay, Aldo M., Francyelle L. Medina, Isaac Jales CS, and Patrice Bertail. "First-Order Integer Valued AR Processes with Zero-Inflated Innovations." In Workshop on Nonstationary Systems and Their Applications, pp. 19-40. Springer, Cham, 2021.
Examples
test <- simul_zinarp(alpha = 0.1, lambda = 1, n = 100)
e.test <- estimate_zinarp(x = test, p = 1, iter = 800, innovation= "Poisson")
alpha_hat <- mean(e.test$alpha)
lambda_hat <- mean(e.test$lambda)
data(slesions)
e.slesions <- estimate_zinarp(slesions$y, p = 1, iter = 800, innovation = 'ZIP')
alpha_hat_slesions <- mean(e.slesions$alpha)
lambda_hat_slesions <- mean(e.slesions$lambda)
rho_hat_slesions <- mean(e.slesions$rho)