setar.sim: Simulation and bootstrap of Threshold Autoregressive model...

In tsDyn: Nonlinear Time Series Models with Regime Switching

Simulation and bootstrap of Threshold Autoregressive model (SETAR)

Description

Simulate or bootstrap a Threshold AR (SETAR) The setar.sim function allows to simulate a SETAR model from scratch. The most important argument is the B argument, which should be a one row matrix, with first the constant/trend arguments, then the slope coefficients, and this for each regime (lower, middle, high). Other arguments such as lag, nthresh indicate the dimension of this matrix. As an example, a SETAR with 2 lags, 1 threshold, a constant, would have coefficient in the order: c(const_L, phi_1_L, phi_2_L, const_H, phi_1_H, phi_2_H) where L is for Lower regime, H for Higher.

setar.boot on the other side resample/bootstraps an existing setar output. It uses a recursive approach, reconstructing the series. Residuals from the original model are resampled using different bootstrap schemes, see resample_vec.

Usage

setar.boot(
  setarObject,
  boot.scheme = c("resample", "resample_block", "wild1", "wild2", "check"),
  seed = NULL,
  ...
)

linear.boot(
  linearObject,
  boot.scheme = c("resample", "resample_block", "wild1", "wild2", "check"),
  seed = NULL,
  ...
)

setar.sim(
  B,
  n = 200,
  lag = 1,
  include = c("const", "trend", "none", "both"),
  nthresh = 1,
  Thresh,
  starting = NULL,
  innov = rnorm(n),
  ...
)

linear.sim(
  B,
  n = 200,
  lag = 1,
  include = c("const", "trend", "none", "both"),
  starting = NULL,
  innov = rnorm(n),
  ...
)

Arguments

setarObject	Bootstrap: the setar object to resample data from.
boot.scheme	Bootstrap: which resampling scheme to use for the residuals. See resample_vec.
seed	Bootstrap: seed used in the resampling
...	additional arguments for the unexported setar.gen.
linearObject	Bootstrap: the linear object to resample data from.
B	Simulation: vector of coefficients to simulate from.
n	Simulation: Number of observations to simulate.
Thresh, nthresh, lag, include	Simulation: parameters for the SETAR to simulate. See setar for their description.
starting	Simulation: Starting values (same length as lag)
innov	Simulation: time series of innovations/residuals.

Value

a list with the simulated/bootstrapped data and the parameter matrix used.

Author(s)

Matthieu Stigler

See Also

SETAR to estimate a SETAR, arima.sim to simulate an ARMA.

Examples

##Simulation of a TAR with 1 threshold
TvarMat <- c(2.9,-0.4,-0.1,-1.5, 0.2,0.3)
sim<-setar.sim(B=TvarMat,lag=2, type="simul", nthresh=1, Thresh=2, starting=c(2.8,2.2))
mean(ifelse(sim>2,1,0))	#approximation of values over the threshold

#check the result
selectSETAR(sim, m=2)

##Bootstrap a TAR with two threshold (three regimes)
sun <- (sqrt(sunspot.year+1)-1)*2
sun_est <- setar(sun, nthresh=2, m=2)
sun_est_boot <- setar.boot(sun_est)
head(sun_est_boot)

##Check the bootstrap: with no resampling, is it the same series?
sun_est_boot <- setar.boot(sun_est, boot.scheme = "check")
all.equal(as.numeric(sun), sun_est_boot)

tsDyn documentation built on Feb. 16, 2023, 6:57 p.m.