A R-Shiny app for visualising shrinkage regression methods

It is known in the statistics literature that the least squares linear regression estimator cannot handle high-dimensional data ($n < p$). A typical solution to this is to use an estimator with shrinkage property, e.g. the Lasso and the Ridge estimator.

This Shiny app will simulate a data with correlated features. The first two columns of the design matrix are generated with coefficient 1 and the rest are 0. By looking at various plots, we can better understand the bahaviour of each of the estimators.

Running the app

You can try using either one of the following three options:

• This is a free Shiny app hosted on shinyapps.io. It has an usage limit so it is not guaranteed to work.

• This app works by deploying this Shiny app through Google Cloud. It also has an usage limit.

• Locally running the code below:

library(shiny)
library(glmnet)
library(tidyverse)
library(directlabels)
library(mvtnorm)

shiny::runGitHub(
    repo = "shrink_shiny", 
    username = "kevinwang09", 
    ref = "bindr",
    subdir = "main")