Multivariate Analysis and Visualization using R Package muvis

TL;DR

The paper introduces the muvis R package, offering comprehensive tools for multivariate data analysis, including novel methods for identifying features with different distributions between groups, demonstrated on NHANES data.

Contribution

The paper presents muvis, an R package that integrates existing multivariate analysis tools with two new methods, VKL and VVKL, for enhanced feature comparison between groups.

Findings

Demonstrated muvis on NHANES dataset.

Implemented two novel divergence methods, VKL and VVKL.

Provided a comprehensive toolkit for multivariate analysis in R.

Abstract

Increased application of multivariate data in many scientific areas has considerably raised the complexity of analysis and interpretation. Although quite a few approaches have been put forward to address this issue, there is still a gap between the most efficient proposed methods and available software. muvis is an R package (core team (2017)) which is a toolkit for analyzing multivariate datasets. Several tools are implemented for common analyses of multivariate datasets, including preprocessing, dimensionality reduction, statistical analysis, Probabilistic Graphical Modeling, hypothesis testing, and visualization. Furthermore, we have implemented two novel methods--Variable-wise Kullback-Leibler Divergence (VKL) and Violating Variable-wise Kullback-Leibler Divergence (VVKL)--which are proposed to find the features with most different probability distributions between two specific groups of samples. The main aim of the package is to provide a wide range of users with different levels of expertise in R with a set of tools for comprehensive analysis of multivariate datasets. We exploited the NHANES dataset to declare the functionality of muvis in practice.