The package offers a method to discretize multivariate continuous data using a grid that captures the joint distribution via preserving clusters in the original data. Joint grid discretization is applicable as a data transformation step before using other methods to infer association, function, or causality without assuming a parametric model.
Most available discretization methods process one variable at a time, such as ‘Ckmeans.1d.dp’. If discretizing each variable independently misses patterns arising from the joint distribution of multiple involved variables, one may benefit from using the joint discretization method in this package.
install.packages("GridOnClusters")See the Examples vignette of the package.