[R-pkgs] clusterMI: Cluster Analysis with Missing Values by Multiple Imputation

[Vincent Audigier]

Dear all,

I am pleased to announce the release of a new package named 'clusterMI' on CRAN.

clusterMI allows clustering of incomplete observations by addressing missing values using multiple imputation.

For achieving this goal, the methodology consists in three steps:

1. missing data imputation using tailored imputation models: four multiple imputation methods are proposed, two are based on joint modelling (JM-GL and JM-DP) and two are fully sequential methods (FCS-homo and FCS-hetero).
2. cluster analysis of imputed data sets: six clustering methods are available (kmeans, pam, clara, hierarchical clustering, fuzzy c-means and gaussian mixture), but custom methods can also be easily used.
3. partition pooling: the set of partitions is aggregated using NMF based method. An associated instability measure is computed by bootstrap. Among applications, this instability measure can be used to choose a number of clusters with missing values.

The package also offers several diagnostic tools for tuning the number of imputed data sets, for checking convergence in sequential imputation, for checking the fit of imputation models, etc.

This is the first version of the package, your feedback and suggestions are welcome!

Please find more details and download the package from the following link:https://cran.r-project.org/package=clusterMI

Best regards,

V. Audigier

-- 
Vincent AUDIGIER
Associate Professor, CNAM
2 rue Conté 75003 Paris
Office 35.3.21
Tel: 01 40 27 27 19
Website:http://vincentaudigier.weebly.com/