mcvis is a R package for visualising multicollinearity in a data
design matrix. The underlying methodology uses resampling techniques to
identify groups of variables that causes multicollinearity.
You can learn more about mcvis from this
vignette.
mcvis can be installed using the devtools package.
devtools::install_github("kevinwang09/mcvis")Using a mcvis bipartite plot, variables (bottom row) that cause strong
collinearity are visualised as bolded lines connecting with our “tau”
statistics (top row).
library(mcvis)
library(ggplot2)
set.seed(1)
p = 10
n = 100
X = matrix(rnorm(n*p), ncol = p)
## Inducing collinearity into the design matrix
X[,1] = X[,2] + rnorm(n, 0, 0.1)
mcvis_result = mcvis(X)
plot(mcvis_result)
- Lin, C., Wang, K. Y. X., & Mueller, S. (2021). mcvis: A new framework for collinearity discovery, diagnostic and visualization. Journal of Computational and Graphical Statistics, 30:1, 125-132.