simulate(model, ...)
## S3 method for class 'ARMA':
simulate(model, y0=NULL, input=NULL, input0=NULL,
start=NULL, freq=NULL, sampleT=100, noise=NULL, sd=1, Cov=NULL,
rng=NULL, noise.model=NULL, compiled=.DSEflags()$COMPILED, ...)
## S3 method for class 'SS':
simulate(model, input=NULL,
start=NULL, freq=NULL,sampleT=100, noise=NULL, sd=1, Cov=NULL,
rng=NULL, compiled=.DSEflags()$COMPILED, ...)
## S3 method for class 'TSestModel':
simulate(model, input=inputData(model),
sd=NULL, Cov=NULL, ...)monoise$w (in that order)
have time series properties, these are given to the output. If noise is not supplied then random values will be generated
using other supplied information or defaults. The rng will be set first if
it is specified.
The default noise generation will be N(0,I)
If Q is not square in a non innovations state space model (i.e. the system
noise has a dimension less than the state dimension), then it is
padded with zeros, so generated noise of higher dimension has no
effect. If sd is supplied, then w as describe below will be N(0,sqr(sd)).
sd can be a vector of p elements corresponding to each of the p
outputs.
If noise is supplied it should
be a list of the necessary noise processes. For non-innovation form state
space models the list must have elements w, e, and w0.
(w0 is w for t=0 in
state space model and prior lags in ARMA models.) For innovation form state
space models and ARMA models with MA components the list should have elements
w and w0, but if w0 is not specified it is set to zero.
For ARMA models with no MA components (i.e. VAR models) the list needs
only w.
In this case, and in the innovations form state space model with w0=0,
a matrix may be supplied in place of a list. w should be
a sampleT by p matrix giving the noise for t=1
to sampleT.
If noise is specified sampleT will be set to the number of periods
in w.
If noise$w0 is a matrix (rather than a vector) for a state space model
simulation (as it is for ARMA simulations) then it is set to a
vector of zeros. This provides compatability with VAR models (ARMA
models with no lags in B).
Input must be specified for ARMA models with model$C not NULL and
state space models with model$G not NULL..
In general ARMA and SS simulations will
not produce exactly the same results because it is impossible to
determine necessary transformation of initial conditions and w0.
makeTSnoise,
TSmodel,
TSdata,
ARMA,
SSmod1 <- ARMA(A=array(c(1,-.25,-.05), c(3,1,1)), B=array(1,c(1,1,1)))
AR <- array(c(1, .5, .3, 0, .2, .1, 0, .2, .05, 1, .5, .3) ,c(3,2,2))
VAR <- ARMA(A=AR, B=diag(1,2))
print(VAR)
simData <- simulate(VAR)
C <- array(c(0.5,0,0,0.2),c(1,2,2))
VARX <- ARMA(A=AR, B=diag(1,2), C=C)
simData <- simulate(VARX, sampleT=150, input=matrix(rnorm(300),150,2))
MA <- array(c(1, .2, 0, .1, 0, 0, 1, .3), c(2,2,2))
ARMA <- ARMA(A=AR, B=MA, C=NULL)
simData <- simulate(ARMA, sampleT=200)
ARMAX <- ARMA(A=AR, B=MA, C=C)
simData <- simulate(ARMAX, sampleT=150, input=matrix(rnorm(300),150,2))
data("eg1.DSE.data.diff", package="dse1")
model <- estVARXls(eg1.DSE.data.diff)
simData <- simulate(model)
ss <- SS(F=array(c(.5, .3, .2, .4), c(2,2)),
H=array(c(1, 0, 0, 1), c(2,2)),
K=array(c(.5, .3, .2, .4), c(2,2)))
print(ss)
simData <- simulate(ss)
testEqual(simData, simulate(ss))
testEqual(simData, simulate(ss, rng=getRNG(simData)))
simData2 <- simulate(ss,
noise=list(w=matrix(runif(300), 150,2), w0=runif(2)))
simData3 <- simulate(ss, noise=matrix(runif(400), 200,2))Run the code above in your browser using DataLab