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.

- Github repository: https://github.com/rlph50/garma
- Added a
`NEWS.md`

file to track changes to the package.