...tbds:
- check and fix PR #16 by ztime: Fix for different levels in training data and x.
- add test for allNAs and kproto_gower
- check NA treatment (argument + remove of NAs, ln 133ff + are they ignored if na.rm=F?)
- migrate to 0.4-1 (after missing, gower and initialization updates)
- maybe add variable importance (LO/perm) to post-validate selection od lambda
- check -itemize environment in validation_kproto.Rd (v0.3-14)
- clprofiles():
1) add legends for factors
2) check additional input parameter for a) w/o ask = T b) specify mfrow/mfcol for simultaneous plots of several variables (+ add error check for this option if # of variables does not match grid)
...is there aby potential impact on plot.kproto?
- try to remember why has there been clprofiles and not plot.kproto so far? (...x argument for keep.data = F? )
clustMixType_0.4-1 2024-06-05 - for type=gower: enable na.rm=imp.onestep and fix na.rm=no
clustMixType_0.4-1 2024-06-05 - enable clustervalidation for type=gower
clustMixType_0.3-17 2024-06-05 - fixed minor issues: print.kproto (num of categoricals); kproto (consistent description of the type argument)
clustMixType_0.3-16 2023-12-17 - kprototypes: new default type = "Huang" added (ln.177) to make sure. (internally relabelled as type = "standard". old call type = "standard" is still possible.)
clustMixType_0.3-15 2023-12-13 - kprototypes: ln. 272-281 removed (missings removed twice for na.rm = "yes", the 2nd time is redundant)
(- kprtotypes_gower: ln. commented 135-163 removed to be the code more readable.)
clustMixType_0.3-14 2023-12-10 - winbuilder note: validation_kproto.Rd: \itemize environment changed: unlike argument lists, \item in these formats is followed by a space and the text (not enclosed in braces)
clustMixType_0.3-13 2023-12-09 - Date field changed to ISO 8601 yyyy-mm-dd format according to Rwinbuilder check note.
- kproto_gower.Rd, \item{lambda} ...\code{code}{type = "gower"}... corrected.
clustMixType_0.3-12 2023-10-25 - implemented three initialization strategies, for which 3 help functions were added
- clean up lambdaest
- include lambdaest into kproto
- fixed some minor typos
- fixed: no output if keep.data = FALSE and nstart>1
clustMixType_0.3-11 2023-09-05 - .rd file for stability_kproto
- bugfix stability_kproto: undefined input parameter
clustMixType_0.3-10 2023-09-01 - cluster stability determination added, along with updated tests
- typos fixed in clustervalidation.R
clustMixType_0.3-9 2022-12-13 - number of test cases for validation_kproto reduced to decrease time for check on CRAN
- fixes in kproto_gower (ln 236ff & 336ff): merging of equal prototypes (one of the for R < 4.1.0 and in ln. 328 the ordinal variables were not considered)
- fix for ordinal variables if k is data frame in kproto_gower (ln. 204ff)
- logo replaced
clustMixType_0.3-8 2022-12-10 - examples from validation_kproto changed to \dontrun{} as they are too slow for CRAN submission
clustMixType_0.3-7 2022-12-09 - fixes according to win-builder: importFrom("stats", "median"), .Rd for na.rm and plot.kproto
- NA treatment for kproto_gower (ln. 424): cluster membership NA if all varianbles are NA. Note kproto_gower does not allow for na.rm = "yes" data with removed missings are passed from kproto().
- validation_kproto (ln. 983)/kproto (ln.920): class(object) != "kproto") replaced by: !inherits(object, "kproto")
clustMixType_0.3-6 2022-12-07 - internal imputation and one step imputation of doi = 10.1007/s00357-022-09422-y implemented
- changing the argument na.rm from logical to character
- backwards compatibility is given, since TRUE is changed to "yes" and FALSE to "no" automatically
- added some new tests in test_basic regarding the imputation
clustMixType_0.3-5 2022-12-06 - fixes in test_basics (prediction for only 1 obs) and validation_kproto (stop for gower).
clustMixType_0.3-4 2022-12-04 - plot.kproto added (wrapper aorund clprofiles).
clustMixType_0.3-3 2022-12-04 - predict.kproto extended to data frame with only one obs. (ln.495, suggestd by G.Terziysky).
- additional $stdization element returned by kproto_gower() for further use in predict function.
- predict.kproto() for type = "gower" changed (now uses standardization from training data.
clustMixType_0.3-2 2022-11-19 - example added to .Rd
- commented chunks removed
- some tests for gower extension added
- error check in validation_kproto() ln. 972f as long as if it is not implemented for type == "gower"
clustMixType_0.3-1 2022-11-13 - kproto_gower.R added (internally called by kproto) (...not built --> v0.3-2)
- impacts on kproto (new type argument), print.kproto and predict.kproto
clustMixType_0.2-15 2021-08-16 - validation_kproto: bug fix for special case of variable selection by lambda for no mixed-type data
- few minor bug fixes for special cases with missing values
clustMixType_0.2-14 2021-07-14 - bug fix of bug fix (ln.202): check for lambda == 0 only if length(lambda) == 1 (otherwise error for is.null(lambda))
clustMixType_0.2-13 2021-07-09 - bug fix for special case of variable selection by lambda for no mixed-type data; validation_kproto will follows shortly
clustMixType_0.2-12 2021-03-10 - after call with uwe also in validation_kproto.Rd \href changed in to \doi (...such that it will be changed with next CRAN submission).
clustMixType_0.2-11 2021-03-09 - \href to \doi changed according to rwinbuilder check note for R devel.
clustMixType_0.2-10 2021-03-01 - bug fix: save kproto object if there is a clusterpartition with same number of cluster but different validation index; comparison of several indices of same size clusters possible
clustMixType_0.2-9 2020-11-04 - update error-messages, if index calculation with k = 1 cluster or an kproto-object with only one cluster.
clustMixType_0.2-8 2020-11-02 - personalized lambda input with validation_kproto added
clustMixType_0.2-7 2020-06-11 - tibble import added to namespace (ln.107 kprototypes.R) and description
clustMixType_0.2-6 2020-05-29 (not built --> v0.2-7)
- enable input of tibbles
- fix par(ask="FALSE") into par(ask=FALSE)
- check done: no impact of NAs in predict or clprofiles
clustMixType_0.2-5 2020-04-22 - fix output for validation indices: k_opt (integer) and index_opt (numeric) instead of named numeric k_opt
clustMixType_0.2-4 2020-04-16 - bug fix ln. 165f
- assign cluster NA to obs. with NAs in all variables
- l. 287 & 330 only update prototypes for variables where not all values are NA
clustMixType_0.2-3 2020-03-23 - implemented wrapper validation-index-functions
- in validation_kproto: added argument kp_obj=c("optimal","all") if output should contain only optimal kproto-object or all calculated kproto-objects
- in validation_kproto: explanation added in help for situation object!=NULL & data!=NULL
- this NEWS file added
clustMixType_0.2-2 2020-03-03 - initial prototypes without NAs
- bug fix: sample in case of multiple minima (which(x==min()) instead of which.min(x))
- bug fix: pass na.rm for nstart > 1
- Example of ValIndices: added kproto(?, keep.data=TRUE) and fix k_{max} instead of k_max
- hexbin added via usethis::use_logo("C:/R/contribs/clustMixType.png")
clustMixType_0.2-1 2019-02-25 - validation indices added (cindex, dunn, gamma, gplus, mcclain, ptbiserial, silhouette, tau)
- Bugfix (case k=1: line 229, following the check for equal cluster; line 285: following the check for empty cluster)
- changing the meaning of verbose and adding some information-outputs
- Simplification of the code-structure (line 211ff early check for equal clusters; line 314ff final update of prototypes and dists)
- Bugfix (occurrence of equal prototypes in last iteration: line 288)
- remove non-sense rownames of protos: line 345
- citation file added
clustMixType_0.1-36 2018-09-01 - testthat error for R3.5.1 and vector valued lambda: as.matrix() added (ln.207/211/250/256/409/415)
clustMixType_0.1-35 2018-08-31 - Several cat() statements changed into messages (ln. 127, 229)
- verbose argument added (prevents counting of NAs)
- stop for data frame with numerics/factors only + reference to kmeans()/kmodes() (ln.114f.)
- warning added for obs. with all variables NA if na.rm = FALSE (ln. 132ff)
- Update .Rd some typos / reformulations.
clustMixType_0.1-34 2018-06-22 - Bugfix (non-scalar k) for NAs / additional checks on k / (row.)names for result added
- Error check nrow(x) < k moved from ln.127 to ln. 152/158/169 (k be integer, vector or data frame)
- Warning added in ln.151 for case k no integer.
- ln.156: Error check: k < 1
- ln.307: names(clusters) <- row.names(dists) <- row.names(x) in case of na.rm=T
- roxygen code changed from Rd summary.kproto into summary (R CMD check NOTE)
clustMixType_0.1-33 2018-06-17 - Bugfixes for NA extension:
- ln.117-120: block with check for NAs moved before NA exclusions
- ln.271: missing na.rm=TRUE added
clustMixType_0.1-32 2018-06-08 - handling of NAs (based on version clustMixType_0.1-30):
- argument na.rm added
- ln.111ff: rows with NAs deleted + message and error check
- vnum/vcat: (ln.128) & lamndaest (ln.564ff) na.rm=T, (ln.128) length(z) by sum(!is.na(z))
- protos for k==1: (ln.175) na.rm=T
- d1: (ln.180 k==1, 224 k>1, 388 predict) rowSums(d1, na.rm=T)
- d2: (ln.184, 229, 309) d2[is.na(d2)] <- FALSE # instead of rowSums(d2, na.rm=T)
*** [clustMixType_0.1-31 ...version skipped (tests on potential Rcpp integration ...postponed)] ***
clustMixType_0.1-30 2017-11-02 kproto(): bugfixes ln.122: anyfact instead of anycat
ln.110 k_input <- k and call for nstart > 1 with k = i_input at ln.292 in case of merged clusters.
clustMixType_0.1-29 2017-10-15 clprofiles(): par(ask=FALSE) added in ln.473
clprofiles(): colours for k<3 fixed
clustMixType_0.1-28 2017-09-04 summary.kproto(): correct cluster IDs in summary tables for numeric variables.
clustMixType_0.1-27 2017-08-30 bugfix kproto after merge of empty clusters (ln.245-251, mail R.Aschenbruck -- error occurred in subsequent check for equal prototypes)
clustMixType_0.1-26 2017-08-30 clprofiles(): col argument added (-> fix colors in bars and boxes),
lambdaest(): num.method and fac.method added (-> sd instead of var according tu huang's 97 paper), thx for feedback added as a comment in *.R
clustMixType_0.1-25 2017-08-16 minor changes in examples .Rd files lambdaest() and summary.kproto() (wrong cluster number k)
clustMixType_0.1-24 2017-08-16 summary.kproto() added; keep.data argument added to kproto()
clustMixType_0.1-23 2017-08-15 lambdaest extension: outtype argument; check: at least one lambda > 0 (ln.92)
clustMixType_0.1-22 2017-07-27 Note R CMD check --as-cran fixed: hidden *.Rhistory file removed from folder
clustMixType_0.1-21 2017-07-27 method = 2 added to lambdaest(); bugfix: error check on length(lambda) moved from ln. 91 ~> ln.101
clustMixType_0.1-20 2017-07-17 allow vor variable specific vector valued lambdas (...if(length(lambda) > 1) d2 <- d2 %*% lambda[catvars])
clustMixType_0.1-19 2017-07-17 extension for k=1: special case, ln.147 ; initialization of clusters/... moved before check for equal prototypes (ln.141)
clustMixType_0.1-18 2017-07-17 k of type data frame added; rm(ids) moved upwards in !is.data.frame(k){}
clustMixType_0.1-17 2016-11-10 Bugfix for single categorical/numeric variables (several dimension drop = FALSE added).
clustMixType_0.1-16 2016-02-27 DOI added for CRAN submission.
clustMixType_0.1-15
clustMixType_0.1-14 2016-02-26 Rename of package.
######### Changelog of former package up to rename into clustMixedType.
kprototypes_0.1-13 2016-02-26 Several importForm ob base packages added accordimg to chges in CRAN policy.
kprototypes_0.1-12 2016-02-25 Reference changed.
roxygen code for export changed (* checking S3 generic/method consistency ... NOTE Found the following apparent S3 methods exported but not registered: kproto.default)
kprototypes_0.1-11 2015-01-26 Speedup of kproto() also added to predict.kproto().
kprototypes_0.1-10 2015-01-26 Slight adaption of clprofiles() (clusids <- ) to kmeans like objects. Examples replaced by more explicative artificial ones.
kprototypes_0.1-9 2014-11-30 Bugfix: currentyl 2 versins of print.kproto() ~> one removed & no assignment of class kproto.
kprototypes_0.1-8 2014-11-27 Description file changed several times according to B.Ripley's reco.
...
kprototypes_0.1-5 2014-11-24 print.kproto() function added for conveniance.
kprototypes_0.1-4 2014-11-23 lambdaest() for constant only; stop() for 1D data frames added.
kprototypes_0.1-3 2014-11-22 speed up kprotos() (by ca. factor 100) in dist calculation (~ln. 130-140)
kprototypes_0.1-2 2014-09-23 rename profile() into clprofiles()
kprototypes_0.1-1 2014-09-23 first built