6/14/2018:
1. Revised the arimax function to fix a bug related to io. Now IO can be specified either by, e.g., io=c(57) or xreg=data.frame(IO.57=seq(x)==57), assuming x is the time series.
2. Fixed a bug in the prewhiten function.
3. 1. Fix a bug in the tar function so that is.constant1 and is.constant2 work properly
4. Fix namespace problems.
11/12/2012:
1. revised the arimax (arima) function by using the atanh transformation of the pacf, as implemented by stats:::arima.
2. Function arimax defaults to stats:arima when io, xtransf and transfer are all missing. This speeds up the calculation.
3. Clear up some useless codes in arimax.
10/26/2012:
1. revised versions of arima and arimax; fix the order of armafilter function so that the AR operators are applied first in order to avoid creating many NAs.
2. Fix the warning problem in tsdiag.Arima and tsdiag.TAR.
9/9/2012:
1. arima and arimax functions temporarily removed from the package; new versions are planned to be released.
2. acf and plot1.acf substantially rewritten to avoid foreign functions are called.
3. arima.boot function rewritten to avoid copyright issues.
4. fitted.arimax changed to fitted.arima (manual accordingly changed)
5. Minor fixes to many functions to avoid warnings from partial argument matching problem.
9/7/2012:
1. Replace the .Internal function cat in tar and tsdiag.TAR by calling cat directly.
7/31/2010:
1. fix some bugs in the help manuals.
2. Fix a typo in harmony.R function:
In the first line,
if (!ts(x) || (2 * m) > frequency(x))
is replaced by
if (!is.ts(x) || (2 * m) > frequency(x))
2/15/08:
fixed periodogram so that I(1/2) is correct, i.e., it equals
the sample spectral density at 1/2, whereas for other frequencies, it
is twice the sample spectral density.
changes made in version 0.96
1. Fixed the output of the acf function within tsdiag.Arima
2. Bug fix for arima and arimax so that ill-conditioned covariates are
properly handled, just as done by the arima function in the stat package.
3. R codes of plot.Arima function revised so that the user can set plotting
options for how the predicted values are plotted.
An example follows.
data(hare)
m1.hare=arima(sqrt(hare),order=c(3,0,0))
plot(m1.hare, n.ahead=25,type='b',xlab='Year',ylab='Sqrt(hare)',pch=19)
abline(h=coef(m1.hare)[names(coef(m1.hare))=='intercept'])
Here, the predicted values are plotted as solid diamonds. If the pch option
is not specified, the default is to plot all observed and predicted values
as open circles.
4. R codes of arima and arimax revised to allow xreg not necessarily a data.frame.
5. The default labelling in season for quarterly data is changed to
1Q, 2Q, 3Q and 4Q. The older version has Q1, Q2, Q3 and Q4.
6. The default of y.name in the armasubsets funciton is changed to "Y".