ar: Convert between parameterizations for the AR(p) model

In andrewzm/INLA: Functions which allow to perform full Bayesian analysis of latent Gaussian models using Integrated Nested Laplace Approximaxion

Description

These functions convert between the AR(p) coefficients phi, the partial autorcorrelation coefficients pacf and the autocorrelation function acf. The phi-parameterization is the same as used for arima-models in R; see ?arima and the parameter-vector a in Details.

Usage

   inla.ar.pacf2phi(pac)
   inla.ar.phi2pacf(phi)
   inla.ar.pacf2acf(pac, lag.max = length(pac)+1L))
   inla.ar.phi2acf(phi, lag.max = length(pac)+1L))

Arguments

pac	The partial autorcorrelation coefficients
phi	The AR(p) parameters phi
lag.max	The maximum lag to compute the ACF for

Value

inla.ar.pacf2phi returns phi for given pacf. inla.ar.phi2pacf returns pac for given phi. inla.ar.phi2acf returns acf for given phi. inla.ar.pacf2acf returns acf for given pacf.

Author(s)

Havard Rue [e-mail]

Examples

 pac = runif(5)
 phi = inla.ar.pacf2phi(pac)
 pac2 = inla.ar.phi2pacf(phi)
 print(paste("Error:", max(abs(pac2-pac))))
 print("Correlation matrix (from pac)")
 print(toeplitz(inla.ar.pacf2acf(pac)))
 print("Correlation matrix (from phi)")
 print(toeplitz(inla.ar.phi2acf(phi)))

andrewzm/INLA documentation built on May 10, 2019, 11:12 a.m.