acf
or
ar
.
acf.plot(z, layout=c(nser,nser), conf.int=T, type="h", ask=F, main=NULL, ci=0.95, ...)
acf
, including components
acf
and
lag
, or a list as returned by the
function
ar
with component
partialacf
.
TRUE
, add a band about the zero line representing an approximate
100
ci
% confidence interval.
"h"
(high density),
this could be
"l"
(lines),
"p"
(points), or
"b"
(both lines and points).
TRUE
, the user will have to hit the return key between pages.
This allows the user to inspect each graph when the graphics device
does not stop automatically between pages.
0.95
.
It is ignored when
conf.int = FALSE
.
conf.int
is
TRUE
, approximate 95% confidence limits about
0
will be
added to the plot.
The
layout
argument is used to
create a matrix of plots, and a title for the entire figure is produced.
The graphical parameters
mfrow
and
mfcol
are respected, that is, the
layout
argument gives the plots that are produced in each frame of the
current
mfrow
(
mfcol
). Other graphical parameters (e.g.
col
,
lty
)
can be given as arguments.
Since the individual plots are drawn using
subplot
, the axis coordinates
are all relative (to a 1 by 1 square); hence, functions like
locator
or
abline
may do strange things. Use
locator()
to
find out what the true coordinates are, if these are needed to add text, lines,
or such.
The confidence interval is computed as
qnorm(0.5+ci/2)/sqrt(n.used)
where
n.used
is the number of nonmissing observations in the time series.
The interval can only be provided when
n.used
is a component of
z
.
Note that this interval is valid only if
z$acf
represents the
correlation function or the
partial autocorrelation function, it is meaningless when
z$acf
is the
covariance function.
The chapter "Analyzing Time Series" of the S-PLUS Guide to Statistical and Mathematical Analysis.
acf.plot(acf.lynx, conf.int=F) acf.plot(bonds.ar <- ar(bonds.yield)) pacf.lynx <- acf(lynx, 36, "partial", plot=F) acf.plot(pacf.lynx, conf.int=T, type="l")