ARMA.var: Covariance/correlation matrix for the stationary ARMA model

In ts.extend: Stationary Gaussian ARMA Processes and Other Time-Series Utilities

Description

This function computes the covariance/correlation matrix for a stationary auto-regressive moving-average (ARMA) model. The user specifies the matrix size n and the function returns a matrix of covariance/correlation values at all times Time[1], ... , Time[n] (in the case where conditioning values are specified using the condvals argument, only the time values for non-conditional values are included). The function requires the model to be stationary, which means that the vector of auto-regression coefficients must give an auto-regressive characteristic polynomial with roots outside the unit circle.

Usage

ARMA.var(
  n,
  condvals = as.numeric(NA),
  ar = numeric(0),
  ma = numeric(0),
  corr = FALSE
)

Arguments

n	Positive integer giving the number of values in the time-series (output variance matrix is an n x n matrix)
condvals	Either a single value NA or a numeric vector with n elements; numeric entries are conditioning values for the generated vector
ar	Vector of auto-regressive coefficients (all roots of AR characteristic polynomial must be outside the unit circle)
ma	Vector of moving-average coefficients
corr	Logical; if TRUE the function returns the correlation matrix; if FALSE the function returns the covariance matrix

ts.extend documentation built on Nov. 15, 2020, 1:06 a.m.