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.
step_orderNorm()to work with parallel processing.
step_best_normalize()to work with parallel processing.
boxcoxin 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
tidymethod to work more generally, provide easy access to chosen transformations (responding to issue 9)
usethisin response to issue 7
n_logit_fitargument, with default of 10000. This should substantially decrease memory use of
orderNormwhile only minimally affecting the out-of-domain approximations.
step_best_normalize, responding to 8
LambertWtransformation types (thank you to Georg M. Goerg, the author of
LambertW, for pointing this out).
center_scaletransform as default when
standardize == TRUE
ggplot2to visualize all transformations.
axefunctionality in order to improve scalability of
tidyfunctionality with bestNormalize and
x.talways matches input vector.
step_orderNormfunctions for implementation within
warn = FALSEwhen calling
bestNormalize. If a transformation doesn’t work, warnings will no longer be shown by default unless
warnis set to
plot.bestNormalizewhich was improperly labeling transformations
exp_xhaving trouble with
standardizeoption, so added option
bestNormalizeto allow a workaround, and changed it so if any infinite values are produced during the transformation, exp_x will not work (that way,
bestNormalizewill not include this in its results).
length(x) > 2000
loofor leave-one-out cross-validation
Added feature to estimate out-of-sample normality statistics in bestNormalize instead of in-sample ones via repeated cross-validation
out_of_sample = FALSEto maintain backward-compatibility with prior versions and set
allow_orderNorm = FALSEas well so that it isn’t automatically selected
Improved extrapolation of the ORQ (orderNorm) method
Added plotting feature for transformation objects
Cleared up some documentation