funpark package in R

Objectives

• Gain experience in package development in R
• Organize study materials as an R package

Background

There are two purposes of this package:

• Archiving study materials more effectively: After studying statistics in the University, I tried to find a means to store and manage what I have learned. I had all of my lecture notes scanned and stored in my OneDrive, but I wanted something more accessible. The size of the entire notes, including the assignments I have done, was almost 4 GB. There had to be a better way to retrive information from the pool I created.
• Learning how to develop an R package: Although using R packages was something I always did on numerous occasions, I wanted to do something more advanced to become more competent in my R skills. I wanted to not only state on my CV/resume that I am familiar with many packages and able to use them, but also show my skills through documentations and prove that I can communicate.

How it works

As of 2020-01-12, funpark package tries to solve two types of tasks:

• Response column(s) transformations: formulating the “response” column of data into another type of response. For example, if a data is prepared to fit a logistic regression (i.e. binary response), then the package provides functions to formulate the response column into either binomial or count response, and vice versa.
• Statistical computations: there are three subjects regarding this task:
  • Computing the fusion estimates, and finding the best tuning parameter using k-fold cross-validation
  • Computing the pmf of \(S = \sum_{i = 1}^{N} X_i\), where \(N \sim \text{Pois}(\lambda)\), \(N\) independent of all \(X_i\)’s, and \(X_i \stackrel{iid}{\sim} p\) for some pmf \(p\) whose support consists of nonnegative integers using pgf and DFT
  • Imputation using the EM algorithm

Click here for the GitHub repository. You can download funpark package by:

# install.packages("devtools")
devtools::install_github("joon3216/funpark")

Learning outcomes

I learned how to:

• create an R package on RStudio
• maintain an up-to-date DESCRIPTION
• write roxygen2 comments in my R scripts to document functions and datasets
• use devtools package and RStudio to maintain an R package

Next steps

• Definitely:
  • Include tests
  • Update the package and extend functionalities
  • Write vignette
• Possibly:
  • Include S3 or R6 methods if applicable
  • Upload on CRAN