gg_raw_pgram().garma() and ggbr_semipara() now
support the periods parameter allowing the user to specify
fixed periods instead of estimating them from the data.Fixes to the garma-package.rd file.
Version 0.9.8 includes an option to stop the automatic generation of
fitted values for the garma() function. This can take a
while if the process is long, so this option may save time during the
model fitting stage.
Version 0.9.7 adds the tsdiag() function for garma
models. The essential white noise test is set to be the Bartletts Tp
test, since this is the only white noise test which has been
theoretically justified on GARMA models. Also the gof()
function has been added - this does the actual work of the Tp test, and
in fact should work and should be valid for normal Arima models as well
as GARMA models.
In the ‘estimation’ process - garma() - there was a bug
with the standard errors which resulted in Nan being returned for some
parameters. This has been fixed. Further fixes have applied to the
Whittle estimation of the standard errors and to the likelihood
calculation.
The fitted values/residuals were not being properly calculated as per 1-step ahead forecasts. This has been rectified (although for non-Gegenbauer ARIMA models there may still be some issues which will be looked at later).
The override of the ggplot function for garma models has been removed
as this was not standard - the correct way to do this is
autoplot() so now the autoplot function has the ability to
generate forecasts and plot them. If no titles and subtitles are
supplied, these routines will now generate some default ones for
you.
A new optimisation method has been added for the garma()
function - this is a genetic algorithm from package GA. It can be used
by specifying opt_method=‘ga’.
The “QML” method of estimation was producing too many errors and appeared to be converging to non-optimal solutions too often. Until such time as this can be fixed, it has been removed as a valid option. In any case it only supports a k=1 model.
Otherwise some redundant code has been removed, and also the restriction that integer differencing be restricted to 1 only - general integer differencing is now supported.
Version 0.9.6 implements new functions to more accurately reflect
residuals, fitted values and predictions. Some changes have been made to
the plotting routines to provide default titles and captions etc.
Further some functions like AIC(), logLik(),
vcov(), and coef() have been implemented, to
provide greater similarity with the standard ‘arima’ functionality.
Finally given the trouble with forecasting with integer differencing > 1, there is a new restriction in the code to restrict the integer differencing to be either 0 or 1. This does not affect the fractional differencing component of the models. If the problems with forecasting higher integer differencing can be resolved, this restriction will be lifted in the future.
In particular please note that the
predict()/forecast() function(s) now use the
algorithm of (2009) Godet, F “Linear prediction of long-range dependent
time series”, ESAIM: PS 13 115-134. DOI: 10.1051/ps:2008015.
WARNING: forecasts generated by this version will be different from previous versions.
NEWS.md file to track changes to the
package.