MonoClus()
: default value fordistmethod
argument is nowNULL
so that the it can be set to Euclidean (whencir.var
is not set) or Gower distance (whencir.var
is set).- Fix typos and a trailing space in the warning message of
as_MonoClust()
function. - Small improvement in how
ncores
argument is processed inperm.test()
andcv.test()
. It did not have a problem before but the improvement helps remove errors in Github Actions CI in Windows environment. \@fig
tags in the vignette is removed.- URL to lifecycle badge in README.md is updated.
- Start unit test component with the first test is for
as_MonoClust()
function.
inertia_calc()
andmedoid()
are now exported functions.is_MonoClust()
to check if an object is an inherited class ofMonoClust
.- Create a S3 function
as_MonoClust()
so other packages (such asPULS
) can reuse print and plot functions. However, no implementation is done in this package.
plot.MonoClust()
allows thecols
argument to have the length greater than the number of leaves. In that case, function will not throw error and only a subset of it will be used.MonoClust.object
gainsalt
object as an indicator of whether an alternate route is available.
- Fix typos.
- More minimum version requirements for other dependencies.
- Remove
bipartvar
column fromframe
object of MonoClust object.var
column should be sufficient for showing splitting variable names. alt
column inframe
object of MonoClust is now a nested tibble containing alternate split details. However, the package does not support specifying an alternate splitting route so users may have to run step-by-step by indicatingnclusters = 2
inMonoClust()
and then on each branch.- Remove
arctic_2019
data set. It was not used anywhere in the examples.
- Add documentation for
MonoClust.object
to explain its structure.
- Min version of dependency
tibble()
is 3.0.0 becausetibble::add_row()
is used with the new behavior. - Fix some typos and clarify some documentation.
- Package is now fully working with all features intended.
- Rewrote a lot of internal functions to clean up unnecessary codes.
- Added ggplot2 versions of CV plot.
- PCP plot for circular data is now named
ggpcp
and uses ggplot2. - New circular add/subtract operators:
%cd+%
,%cd-%
(in degree),%cr+%
, and%cr-%
(in radian). - Added
wind_sensit_2008
data set. - Added a vignette using R Markdown.
- Updated documentation.
MonoClust()
removedperm.test
andalpha
arguments. Users now need to runperm.test()
separately to perform permutation test on a MonoClust.- Removed
labels
,corders
,ran
from the output. Membership
andDist
outputs ofMonoClust()
are nowmembership
anddist
, respectively.
- Removed
plot.MonoClust()
addeduniform
,branch
,minbranch
,stats
,cols.type
, andshow.pval
.cv.test()
now returns unified output for both LOOCV and k-fold. It includes MSE and SE table and a note with the type of cross-validation.abbrev
argument inplot.MonoClust()
andprint.MonoClust()
now accepts descriptive options"no"
,"short"
,"abbreviate"
.method
argument inperm.test()
now uses descriptive options"sw"
,"rl"
and"rn"
. Users can now decide whether to apply Bonferroni correction withbon.adj
argument.predict.MonoClust()
removedna.action
.type
argument now accepts more meaningful options"centroid"
or"medoids"
.- Added parallel processing capability to
perm.test()
andcv.test()
.
- Removed the categorical variable feature to make sure it works well for all other cases.
- Added support for
foreach
when searching for the best split. - Added a
NEWS.md
file to track changes to the package.