tpq/amalgam: Amalgamation-Based Dimension Reduction of Compositional Data

It is difficult to visualize high-dimensional compositional data. One approach involves amalgamating the parts into a new set of composite features called "amalgams". This package uses genetic algorithms to find an optimal set of amalgams that satisfies an objective function. Although amalgams themselves are low-dimensional compositions, they are not sub-compositions. Rather, amalgamation acts like a kind of non-linear dimension reduction that can facilitate data understanding. Note that inferences made on amalgamations may not agree with inferences made on the full composition.

install.packages("remotes")
remotes::install_github("tpq/amalgam")