CRAN release: 2023-10-07
- Internal change to check package version with character instead of numeric.
- Internal minor change to in_disaggr (incohesive parameters)
- A more informative error message if one tries to differentiate a time-series with only one observation
- A more informative error message if the rank becomes imperfect after decorrelation
CRAN release: 2023-02-20
- Breaking change : The order of the output of the
reViewOutputobject has been reversed. reView and rePort can now take a language object as name arguments and return language objects inside their name attributes (PR #86)
getOption("digits")for the benchmarks and
max(3L, getOption("digits"))for the praislm objects.
CRAN release: 2022-12-13
- Internal change from ggplot2 size argument, which is deprecated for lines, to linewidth (PR #78)
CRAN release: 2022-02-22
- Added the signature
c("disaggR","missing")for Ops group generic. (PR #54)
- Inner calls to aggregate are now redirected to a faster non-exported function. (PR #55)
- Estimation spans and outliers are now handled in preset models in rePort and reView (PR #57)
- NULL labels are now removed even outside the plot margins (PR #59)
in_scatterfunction now substracts the outliers contributions from the low-frequency serie before computing the in_scatter comparison. (PR #63)
in_scatterhas now an additional arguments :
type(as every in_ function). For now, the only use for the type argument is to allow changes scatterplots for levels models. (PR #63)
- The smoothed.part, for differencied benchmarks, has been set to a new base (its aggregated value is 0 in 2000). That way,
reUseBenchmarkis fixed if used on high-frequency series that have a different start than the previous one. This has no impact on the benchmarked serie. (PR #65)
CRAN release: 2021-08-23
- New vignette : Introduction to disaggR.
- The error of
typearguments has been changed, because it didn’t include “contributions”.
cex.mainparameters now overwrite their default if used inside
- By default, in the
autoplotmethods, the axis annotations are now automatically set to be smaller if needed.
CRAN release: 2021-07-21
- Switched ggplot2, rmarkdown and shiny from imports to suggests. RColorBrewer has replaced scales as an import. disaggR can now be installed with far less dependencies. Hence, the
autoplotgeneric is not reexported anymore. ggplot2 has to be attached to allow the use of
ggplot2::prefix, by example with
CRAN release: 2021-06-18
- added support of outliers.
twoStepsBenchmark, the set.coeff names used to be replaced by
NCOL(hfserie) == 1Land
length(set.coeff) == 1L. This behavior was contradictory with the documentation if
set.coeff=c(constant=1L). As for now, set.coeff names will never be replaced. Then, it makes the controls stricter because
set.coeff=c(x.name.herited.from.anywhere=1)will lead to an error.
- as for the
intime-series plots, the y window now ignores the infinite values.
- reView : a shiny reviewing application for twoStepsBenchmarks.
- rePort : a rmarkdown html report for twoStepsBenchmark and reView outputs.
- start.domain and end.domain know crops the hfserie after having calculated the coefficients. That way, it is possible to evaluate the coefficients on a full hfserie, cropping them for the application.
in_samplenow generates a more general class
"tscomparison", with a
"in_sample"func attribute. In previous versions, the S3 class was named “insample”.
- the new functions
in_revisionsalso produce tscomparisons, with plot and autoplot methods.
- the graphics are prettier thanks to the package scales.
- the plot and autoplot methods now have xlab, ylab, start, end, col, lty, show.legend, main and mar arguments. The autoplot methods have also a theme argument.
- removed the c++ code to improve readability.
- most of the stats methods for time-series now coerce twoStepsBenchmarks or rateSmooths into time-series.
reUseBenchmarknow induces a
set.smoothed.partelement in model.list if
threeRuleSmoothmakes it easier than bflSmooth to procede to a rate smooth.
CRAN release: 2020-10-09
- various optimizations including cache for bflSmooth, which is now much faster, and alternative internal methods for time-series.
- added a weights arg to bflSmooth, that reproduces the lissage en taux methodology.
- the praislm and twoStepsBenchmark summaries now print some disclaimer to tell if the regression includes a differenciation.