SimpleLinReg.jl

My first ever Julia package for doing simple linear regression in two variables.
Author

Debartha Paul

Published

February 1, 2023

GitHub

Introduction

This package is a result of my preparation for the Google Summer of Code, 2023. Unfortunately, I couldn’t make it to the GSoC 2023, but I did create this package. I build this package from the groundup using Julia. The package currently consists of 1 function which enables anyone to find the simple linear regression between two variables.

Installation

This package is not intended for the Julia general package repository so we have to install it using the GitHub address.

  • Step 1- Open Julia REPL in your device. Then run the following command:
using Pkg
Pkg.add(url = "https://github.com/itsdebartha/SimpleLinReg.jl.git")
  • Step 2- The package will be installed. To check, press ] which will open the Pkg REPL. Then run st. You’ll be seeing something like:
(@v1.9) pkg> st
Status `C:\Users\user\.julia\environments\v1.9\Project.toml`
  [4c88cf16] Aqua v0.8.4
  [77b51b56] AverageShiftedHistograms v0.8.9
  [6e4b80f9] BenchmarkTools v1.4.0
  [336ed68f] CSV v0.10.12
  [13f3f980] CairoMakie v0.11.5
  .
  .
  .

If you can see SimpleLinReg listed (which you should), then the installation is fine.

Usage

The main function of interest is linreg. It has the basic syntax as:

linreg(x::Vector,y::Vector;intercept::Bool = true)

where x is the regressor and y is the regressand. The intercept field accepts a boolean value which determines whether the intercept term in the simple linear regression be present or not. In short, with intercept=true (by default), we estimate \(\alpha\) and \(\beta\) as in the equation: \[ y=\alpha + \beta x + \epsilon \]

In case of intercept=false, we estimate only \beta as in: \[ y=\beta x + \epsilon \]

Please note that in both the instances, we assume \(\epsilon\) to follow \(N\left(0, \sigma^2\right)\) identically and independently.

Output

The output will have the format as:

SIMPLE LINEAR REGRESSION IN TWO VARIABLES:
==========================================
Slope:       5.04638
Intercept:   1.98306

Help Files

Just like any other Julia package, if you press ? in the REPL, then the Help REPL will open up. Typing linreg will give a description of the function along with the method and its syntax.

Scope of Improvement

There are many subtle improvements that are needed to mark this package as “complete”. Among those are the ability to handle missing values. A more preliminary test of normality is also encouraged. Further, I have planned to implement the significance tests for the estimated parameters.

If you have any suggestions, feel free to contact me. Even better, you could also create an issue with your suggestion in GitHub.