R/dS2_nlin_test.R

In ChrisKust/rexpar: Simplicial Depth for Explosive Autoregressive Processes

dS2_nlin_test <- function(thetaN, alpha, y, exact = FALSE)
{
  dS2 <- rexpar::dS2_lin2(theta = thetaN, y = y, model = "nlinAR1")
  NdS2 <- sqrt(floor((length(y) - 1) / 2) - 1) * (dS2 - 1 / 4) / sqrt(3 / 16)
  if(exact)
  {
    deci <- ((dS2 * (floor((length(y) - 1) / 2) - 1)) < qbinom(alpha, size = (floor((length(y) - 1) / 2) - 1), prob = 1 / 4))
  }
  else
  {
    deci <- (NdS2 < qnorm(alpha))
  }
  list(TS = NdS2, phi = deci)
}