Simulate a realization from a fitted AR model. This is useful in the
parametric bootstrap. Generic function for "Boot" method.
Usage
## S3 method for class 'FitFGN'
Boot(obj, R = 1, ...)
Arguments
obj
the output from FitAR
R
number of bootstrap replications
...
optional arguments
Details
The method of Davies and Harte (1987) is used if it is applicable, otherwise
the Durbin-Levinsion recursion is used.
Value
If R=1, a simulated time series with the same length as the original fitted time series
is produced. Otherwise if R>1, a matrix with R columns and number of rows
equal to the length of the series containing R replications of the bootstrap.
Author(s)
A.I. McLeod
References
McLeod, A.I., Yu, Hao, Krougly, Zinovi L. (2007).
Algorithms for Linear Time Series Analysis,
Journal of Statistical Software.
See Also
SimulateFGN,
DHSimulateDLSimulate
Examples
#Example 1
#Fit a FGN model and determine the bootstrap sd of H
#Measure cpu time. With R=250, it takes about 23 sec
#on 3.6 GHz Pentium IV.
## Not run:
data(NileMin)
outNileMin<-FitFGN(NileMin)
start<-proc.time()[1]
R<-25
Hs<-numeric(R)
Z<-Boot(outNileMin, R=R)
for (i in 1:R)
Hs[i]<-GetFitFGN(Z[,i])$H
BootSD<-sd(Hs) #this is the bootstrap sd
end<-proc.time()[1]
totTim<-end-start
## End(Not run)