Description
Clustering Method Based on Boxplot Statistics.
Description
Following Arroyo-Maté-Roque (2006), the function calculates the distance between rows or columns of the dataset using the generalized Minkowski metric as described by Ichino-Yaguchi (1994). The distance measure gives more weight to differences between quartiles than to differences between extremes, making it less sensitive to outliers. Further,the function calculates the silhouette width (Rousseeuw 1987) for different numbers of clusters and selects the number of clusters that maximizes the average silhouette width, unless a specific number of clusters is provided by the user. The approach implemented in this package is based on the following publications: Rousseeuw (1987) <doi:10.1016/0377-0427(87)90125-7>; Ichino-Yaguchi (1994) <doi:10.1109/21.286391>; Arroyo-Maté-Roque (2006) <doi:10.1007/3-540-34416-0_7>.
README.md
boxplotcluster 0.3
vers 0.2
- Option added for the selection of the units to be clustered (columns or rows);
- A copy of the input dataset is now returned; information is appended in order to store the rows or columns cluster membership;
- Info about the silhouette statistics now returned by the function;
- Minor internal optimisations;
- Improvements and updates to the help documentation;
- New example added to the help documentation;
- Link to the package's vignette added to the help documentation.
vers 0.1 first release to CRAN.