rm(list = ls())
library(tseries)
## Registered S3 method overwritten by 'quantmod':
## method from
## as.zoo.data.frame zoo
library(forecast)
library(fGarch)
## Loading required package: timeDate
## Loading required package: timeSeries
## Loading required package: fBasics
set.seed(3)
bagging = function(series ,h){
model = forecast::Arima(series, order = c(2,0,0),)
fcast = forecast::forecast(model, h = h)$mean
return(fcast)
}
# AR(2) -------------------------------------------------------------------
n = 100 # 50 #100 # 300
nsim = 100
ar = c(.5,.45) # stationary, unit roottei oiroltsoo
h = 1 # 6, 12
h6 = 6
h12 = 12
order = c(2,0,0) ### table eer report hiih
# e <- rnorm(n) # sd up or down 100,15,1
train = vector(mode = 'list', length = nsim)
real = vector(mode = 'list', length = h)
for(i in 1:nsim){
ar.sim = arima.sim( list(order= order, ar = ar),innov = rnorm(n + h)
,n = n + h )
train[[i]] = head( ar.sim, n)
real[[i]] = tail(ar.sim, h)}
nb = 100
va = lapply(train , tsbootstrap ,nb = nb, b = round( n/3),
type = 'block', statistic = bagging, h= h)
bag = c()
org = c()
rel = c()
for( i in 1:nsim ){
bag[i] = mean(va[[i]]$statistic)
org[i] = va[[i]]$orig.statistic
rel[i] = real[[i]] }
MSE_bag = mean ((rel - bag)^2 )
MSE_orig = mean ((rel - org)^2 )
MSE_bag < MSE_orig
## [1] FALSE
knitr::kable(cbind(h = h,nsim, nb , MSE_orig, MSE_bag))
| 1 |
100 |
100 |
1.357414 |
8.668177 |
# TAR(2) ------------------------------------------------------------------
### y[t-1] > g == y[t] = a0 + a1 * y[t-1] + a2 * y[t-2] + e[t]
### y[t-1]<= g == y[t] = b0 + b1 * y[t-1] + b2 * y[t-2] + e[t]
tar2.sim = function(y0, e , n, p1, p2, th){
y = rep(0 ,n)
a0 = p1[1]; a1 = p1[2]; a2 = p1[3]
b0 = p2[1]; b1 = p2[2]; b2 = p2[3]
for(t in 3:n){
if(y[t-1] > th) y[t] = a0 + a1 * y[t-1] + a2 * y[t-2] + e[t]
else y[t] = b0 + b1* y[t-1] + b2 * y[t-2] + e[t]
}
return(y)
}
n = 100
nsim = 100
h=1
train = vector(mode = 'list', length = nsim)
real = rep(NA , nsim)
for(i in 1:nsim){
tar.sim = tar2.sim(0, rnorm(n = n + h), n = n+1,
c(0.1,.09,.07),
c(0.2 , .05 , .04), th = 0)
train[[i]] = head( tar.sim, n)
real[i] = tail(tar.sim,h) }
nb = 100
va = lapply(train , tsbootstrap ,nb = nb, b = round( n/3),
type = 'block', statistic = bagging, h= h)
bag = c()
org = c()
rel = c()
for( i in 1:nsim ){
bag[i] = mean(va[[i]]$statistic)
org[i] = va[[i]]$orig.statistic
rel[i] = real[i] }
MSE_bag = mean ( (rel - bag)^2 )
MSE_orig = mean( (rel - org)^2 )
MSE_bag < MSE_orig
## [1] FALSE
knitr::kable(cbind(h = h,nsim, nb , MSE_orig, MSE_bag))
| 1 |
100 |
100 |
1.280624 |
1.284359 |
# nb = 10
# va = lapply(train , tsbootstrap ,nb = nb, b = round( n/3),
# type = 'block', statistic = bagging)
#
# ba = rep(NA, nsim)
# or = rep(NA, nsim)
# for( i in 1:nsim){
# ba[i] = mean( va[[i]]$statistic )
# or[i] = va[[i]]$orig.statistic
# }
#
# MSE_bag = mean( (real - ba)^2 )
# MSE_orig = mean( (real - or)^2 )
#
# MSE_bag < MSE_orig
#
# knitr::kable(cbind(nsim, nb , MSE_orig, MSE_bag))
# GARCH(2,2) --------------------------------------------------------------
spec = garchSpec(model = list(alpha = c(0.12, 0.04),
beta = c(0.08,.05)),
cond.dist = 'norm')
train = vector(mode = 'list', length = nsim)
real = rep(NA , nsim)
for(i in 1:nsim){
garch.sim = garchSim(spec, n = n + 1, extended = F)$garch
train[[i]] = head( tar.sim, n)
real[i] = tail(tar.sim,h) }
nb = 100
va = lapply(train , tsbootstrap ,nb = nb, b = round( n/3),
type = 'block', statistic = bagging, h= h)
bag = c()
org = c()
rel = c()
for( i in 1:nsim ){
bag[i] = mean(va[[i]]$statistic)
org[i] = va[[i]]$orig.statistic
rel[i] = real[i] }
MSE_bag = mean ( (rel - bag)^2 )
MSE_orig = mean( (rel - org)^2 )
MSE_bag < MSE_orig
## [1] TRUE
knitr::kable(cbind(h = h,nsim, nb , MSE_orig, MSE_bag))
| 1 |
100 |
100 |
0.0863449 |
0.0626598 |