Reputation:
I want to simulate from an ARMA(1,2)
series. I need to initialize $X0$ = 9
. How do I do it in R? I cannot find any provision for initialization in arima.sim
Upvotes: 2
Views: 3456
Reputation: 1049
You may get it using the right combination of the arguments start.innov
, n.start
and innov
that are passed to stats:arima.sim
.
For an AR process with parameter phi
, it is easy to see that the following yields the first simulated value equal to $X_0=9$.
n <- 50
phi <- 0.5
x0 <- 9
set.seed(123)
arima.sim(n, model=list(ar=phi), start.innov=x0/phi, n.start=1,
innov=c(0,rnorm(n-1)))
# [1] 9.00000000 3.93952435 1.73958469 2.42850066 1.28475872 0.77166710
# [7] 2.10089853 1.51136547 -0.50937850 -0.94154210 -0.91643302 0.76586529
# [13] 0.74274647 0.77214469 0.49675506 -0.30746361 1.63318133 1.31444115
# ...
For the ARMA(1,2) you should do the maths, for example I found the following combination for the parameters used below:
set.seed(123)
arima.sim(n=50, model=list(ar=0.5, ma=c(0.4, 0.2)), start.innov=c(0,0,10),
n.start=3, innov=c(rep(0,3),rnorm(47)))
# [1] 9.00000000 6.50000000 3.25000000 1.06452435 0.07789443 1.39348940
# [7] 1.34470092 1.14158321 2.35167337 2.34863643 0.43663646 -0.88237587
# [13] -1.41460329 0.20114479 0.86088655 1.21995661 0.95293237 0.04505243
# ...
Upvotes: 4