R/f.parzen.R

In AmurG/nse: Numerical Standard Errors Computation in R

# Estimation of the generalized derivative of spectral density 
# at the origin with the Parzen kernel
f.parzen = function(g,st,q) {

  size = length(g)

  if (q == 0) {
    spz = (t(g) %*% g)/size
  } else {
    spz = 0
  }

  if (st <= size-1){
    maxlagpz = st
  } else {
    maxlagpz = size-1
  }

  jpz = 1
  while (jpz < maxlagpz) {
    xpz  = jpz / maxlagpz
    s0pz = (t(g[1:(size-jpz)]) %*% g[(1+jpz):size])/size

    if (xpz <= 1/2) {
      wpz = 1-1-6*xpz^2+6*xpz^3
    } else if(xpz <= 1){
      wpz = 2*(1-xpz)^3
    } else {
      wpz = 0
    }

    spz = spz+wpz*(jpz^q)*(s0pz+s0pz)

    jpz = jpz + 1
  }
  return(spz)
}