R/rma.R

In freqdom: Frequency Domain Analysis for Multivariate Time Series

#' Generates \eqn{n} observations of a \eqn{d}-dimensional moving average process with operators A, i.e.
#' \deqn{ Y_t = \sum_{i \in L_A} A_i X_t },
#' where L_A is a set of lags on which operators A_i are defined.
#'
#' @title Moving avarege process
#' @param n number of observations to generate
#' @param d number of dimensions of the process
#' @param A time domain object describing operators 
#' @param noise the underlying X process
#' @export
rma = function(n, d, A, noise=NULL)
{
  lag = rev(A$lags)[1] - A$lags[1]
  X = rar(n+lag,d=d,Psi = matrix(0,d,d), noise=noise)
  linproc(X[lag + 1:n,], A, noise=function(n){ rep(0,n)})
}

rma.old = function(n, lag=2, d=NULL, noise=NULL)
{
  rma.proc(rar(n+lag-1,d=d,noise=noise))
}

# @export
rma.proc = function(TS, lag=2){
  n = dim(TS)[1]
  RES = c()
  for (x in 1:(n-lag+1)){
    new = TS[x + 0:(lag-1),]
    RES = rbind(RES,colMeans(new))
  }
  RES
}