GetFitARMA {FitARMA} | R Documentation |
The algorithm of McLeod and Zhang (2007) is used.
GetFitARMA(y, p, q, pApprox = 30, init = 0)
y |
time series |
p |
AR order |
q |
MA order |
pApprox |
initial AR approximation |
init |
initial parameter estimates |
See McLeod and Zhang (2006).
For AR(1), exact solution obtained via solving a cubic equation.
For MA(1), optimize
is used instead of optim
.
For valid likelihood computation, we need 2*pApprox >= length(y). So if the supplied
value of pApprox does not satify this pApprox is set to length(n)/2.
If pApprox is less than 30, a warning message is displayed.
loglikelihood |
value of maximized loglikelihood |
phiHat |
estimated phi parameters |
thetaHat |
estimated theta parameters |
convergence |
result from optim if used. Otherwise 0. |
algorithm |
"L-BFGS-B" or "Nelder-Mead" or "optimize" or "cubic" |
A.I. McLeod, aimcleod@uwo.ca
McLeod, A.I. and Zhang, Y. (2007, tentatively accepted). Faster ARMA Maximum Likelihood Estimation. Computational Statistics and Data Analysis.
data(SeriesA) z<-SeriesA-mean(SeriesA) GetFitARMA(z, 1, 1) w<-diff.ts(z, differences=1) GetFitARMA(w, 0, 1)