either a numeric vector or a list with components 'ar' and 'ma'. If
'object' is numeric, it is interpreted as a model with no 'ma' part.
lag.max
the maximum number of lags for which to calculate the ACF or PACF
pacf
logical. Should the partial autocorrelations be returned?
plot
logical. Should the ACF (or PACF) be plotted?
xlab, ylab, ylim, type
arguments for 'plot'
complex.eps
a small positive number used to identify complex conjugates: Let
'roots' = the vector of p roots of the characteristic polynomial of
the autoregressive part of 'object'. This is used by
'findConjugates': x[i] and x[j] are considered conjugate if their
relative difference exceeds complex.eps but the relative difference
of their conjugates is less than complex.eps.
We use 'solve' in the 'polynom' package, because it was substially
more accurate for cases we tested in R 2.6.0 than 'polyroot'.
...
optional arguments passed to 'plot'
Details
1. Compute and test stationarity. An ARMA process is stationary if
all the roots of its AR component lie inside the unit circle (Box and
Jenkins, 1970). If the process is not stationary, a warning is
issued, and no plot is produced.
2. Compute and plot the theoretical ACF.
3. Analyze periodicity of any complex roots
Value
a list with the following components
roots
a complex vector of the roots sorted by modulus and sign of the
immaginary part.
acf, pacf
a named numeric vector of the estimated ACF (or PACF of 'pacf =
TRUE').
periodicity
a data.frame with one row for each complex conjugate pair of roots
and columns 'damping' and 'period'.
George E. P. Box and Gwilym M. Jenkins (1970) Time Series Analysis,
Forecasting and Control (Holden-Day, sec. 3.4.1. Stationarity and
invertibility properties of Mixed autoregressive-moving average
processes)
Ruey Tsay (2005) Analysis of Financial Time Series, 2nd ed. (Wiley,
ch. 2)