Breaking changes: - Harmonisation of the names of the functions. Now mostly snake case, e.g. makeDummy() is changed to make_dummy(). - Harmonisation of the names of the options. Now mostly snake case. - Some minor functions where moved to internal functions, and can only be accessed via dsa:::.function.name, e.g. dsa:::.drop31. Full list: .drop31, .fill31, .fill_up, .Add, .day_split, .df2HTML. - dow, dom and day_of_month functions are not included in the package anymore. - The option "Diff" has been removed from dsa(). If the differenced series shall be adjusted, the differencing needs to occur outside of dsa(). - The returned object by dsa() contains sets of xts time series. A few of these have been renamed, especially those contained in sa_result and sa_result2.
Major changes: - vignette: A full vignette has been written with examples from Ollech (2021). - dsa(): Optimised the (optional) mean-correction of the seasonal and calendar components. The options s.window1, s.window2 and s.window3 can now be set to NULL so that the respective step of the dsa procedure is skipped. - outlier(): Improved outlier detection, fixed bug wrt the estimated t-value. - output(): Added seasonality tests, added plot of monthly aggregates, added acf and pacf plots for regression model, improved description. - Inclusion of examplary data. - Inclusion of holiday regressors.
Further changes: - print(): Defined the print behaviour for objects of class daily. - multi_xts2ts(): New function for multiple conversion of time series of class xts to ts. Helpful for working with calendar regressors. - del_names(): New function for working with xts in the context of calendar regressors. It is helpful, because the default behaviour of the xts names can lead to smaller problems when we are combining several xts time series. - dsa(): Time zones of the outputted objects are now fixed by appyling as.Date() when necessary. - dsa(): Now, dsa() correctly handles series that start with one or multiple NA - Using a model_span interferred with the mean_correction. This has been changed. - The package no longer depends on package extrafont.
Bug fixes: - In output: The outlierplot was misleading. Now the major gridlines instead of minor gridlines are drawn to make sure that the dates on the x-axis are helpful.
Changes: - Added options to plot_spectrum
Changes: - Improved outlier detection and adjustment. Set cval higher, if too many outliers are found! - s.window2 can now be set to NULL - Warning due to regressors / the forecast package now avoided
Changes: - Bug fix: Now multiple external regressors can be added correctly - fixed bug in xts2ts for leapyears
Changes: - dow_dummy now has an option delete 29_2 to make it possible to delete the values for the 29th of February - Improved treatment of 29th February
Changes: - In drop31, more iterations are allowed, to find the correct length of the dates for the new variables
Changes: - In output: for the graphics cairo-png is no longer used, as on some devices it does not create graphics
Changes: - Bug fixes: - Some minor code clean-up
Additional Functionality: - get_sa, get_trend, get_original: Inclusion of helper function to quickly get the most important series from the seasonal adjustment
Changes: - output(): Now contains additional visualizations of the outlier adjustment
Changes:
Changes:
Changes:
Changes:
Changes:
Changes:
Changes:
Changes: