o1iv3r / clustimpute Goto Github PK

dummy coding -> numerical values -> value spaces stays {0,1} since one only samples from the original distribution

true support of categorical variables?
Other distance measures (e.g. Gower)

• For example, a recipe step for tidymodels
• Could be part of code, or in a vignette to show the method

Clustering:

• ClusterR::KMeans_arma supports multiple threads

Random sampling:

• The random sampling over the columns could be easily performed in parallel using foreach
• nested foreach: https://cran.r-project.org/web/packages/foreach/vignettes/nested.html

See
https://github.com/ropensci-review-tools/autotest

Calculate and improve

https://github.com/r-lib/covr

Hello, I am trying to plot my cluster assignment and saw this example code on your documentation:

ggplot2::ggplot(res$complete_data,ggplot2::aes(x,y,color=factor(res$clusters))) +
ggplot2::geom_point()

I am doing a cluster analysis with 4 variables and have some missing data. Is there a way to plot the cluster assignments with 4 variables? (i.e., I can't just put x and y as the axes because I have additional variables). I saw that other packages, like factoextra, use PCA to create their cluster plot, but I can't do the PCA with missing data).

I am looking for something along the lines of the attached image. Thanks!

Hi Oliver,

How about adding a new option to support multiple threads so that the missing value imputation could be much faster?

Shicheng

Hi Oliver,
Just a quick comment.

I came across your nice ClustImpute package. Long ago I worked on a problem that is related to your approach: https://stefvanbuuren.name/publications/Imputation%20categorical%20-%20Psychometrika%201992.pdf Perhaps you could take a look at it to see if it is of any use.

Best wishes, Stef.

Background: Constant issues with TravisCI

Issue: Used by the copula package