lambert type transforms again.lambert of type “hh” as option when
allow_lambert_h is set to TRUE, per request
from issue
24dplyr::progress_estimated
(#26)bestLogConstant, that uses the same
machinery to pick the best value of a constant to use when logging a
variable, e.g. the one that makes the distribution look the most normal,
especially useful for non-positive or zero-inflated data. Currently
experimental.step_orderNorm() to work with
parallel processing.step_best_normalize() to work
with parallel processing.boxcox in response to issue
10; thank you to Krzysztof Dyba (kadyb) for the suggestions.yeojohnson, thanks to Emil
Hvitfeldt (EmilHvitfeldt)
for his work on this problem for the recipes package here.tidy method to work more generally,
provide easy access to chosen transformations (responding to issue
9)usethis in response to issue
7n_logit_fit argument, with default of 10000. This should
substantially decrease memory use of orderNorm while only
minimally affecting the out-of-domain approximations.step_bestNormalize to
step_best_normalize, responding to 8LambertW
transformation types (thank you to Georg M. Goerg, the author of
LambertW, for pointing this out).center_scale transform as default when
standardize == TRUET and F to TRUE
and FALSEscales and
ggplot2 to visualize all transformations.butcher and axe functionality in
order to improve scalability of step_* functionstidy functionality with bestNormalize and
step_best_normalizebestNormalizestandardize option from
no_transform so x.t always matches input
vector.step_bestNormalize and
step_orderNorm functions for implementation within
recipes.warn = FALSE when calling
bestNormalize. If a transformation doesn’t work, warnings
will no longer be shown by default unless warn is
set to TRUE.plot.bestNormalize which was
improperly labeling transformationsexp_x having trouble with standardize
option, so added option allow_exp_x to
bestNormalize to allow a workaround, and changed it so if
any infinite values are produced during the transformation, exp_x will
not work (that way, bestNormalize will not include this in
its results).quiet is
FALSE and length(x) > 2000loo for leave-one-out cross-validationbestNormalize function via allow_lambert_h
argument.Added feature to estimate out-of-sample normality statistics in bestNormalize instead of in-sample ones via repeated cross-validation
out_of_sample = FALSE to maintain
backward-compatibility with prior versions and set
allow_orderNorm = FALSE as well so that it isn’t
automatically selectedImproved extrapolation of the ORQ (orderNorm) method
Added plotting feature for transformation objects
Cleared up some documentation