hubEnsembles.Rmd
article now explains how to ensemble samples usinglinear_pool()
linear_pool()
supports requesting a subset of component model sample forecasts when ensembling samples (#144)linear_pool()
supports the specification of the compound task ID set, so that trajectory samples can be correctly ensembled (#144)linear_pool()
supports the simplest case of ensembling samples, where all component samples are collected and returned (#109)
simple_ensemble()
now usesidentical()
to avoid triggering anall.equal.environment()
error. This error would sometimes occur when providing theagg_fun
argument with a custom function. (#134)
- README now points to hubverse R-universe
- Package submission to CRAN
hubEnsembles.Rmd
vignette is now an articlelinear_pool()
now properly splits its pools (#128)linear_pool_quantile()
uses internal package functions only, notHmisc-utils
functions- Functions using
all_of()
are updated to avoid throwing dplyr warnings
- Base R 4.1 pipe (
|>
) is used in place of magrittr pipe (%>%
) - Function examples are simplified
simple_ensemble()
now produces valid distributions for all weighted medians (#122)
- Validate that
weights
argument doesn't contain weights dependent on output type ID for PMF and CDF forecasts (#35)
- Functions now use
map()
andlist_rbind()
in conjunction to avoid superseded warnings from purrr (#117) - Functions now use double quotes or
.data[[]]
as appropriate within dplyr functions to avoid warnings (#117)
- Organization name has been changed to "hubverse-org" (#115)
hubEnsembles.Rmd
vignette now better reflects package capabilities (#29, #113)- Example data that is out of date has been removed (#113)
- Hmisc dependency has been removed (#55)
- hubUtils dependency has been bumped to 0.0.1 or higher, after its split into hubUtils and hubData (#98)
- Roxygen is bumped to 7.3.1
- Lint workflow have been added (#96, #98)
- GitHub workflows have been upgraded (#96, #98)
- Example data has been added (#95)
- Package docs are upgraded to hubStyle theme (#93)
linear_pool_quantile()
now coerces quantile levels to numeric to prevent distfromq errors (#58, #63)
- Initial Release.