How to Plot the Linear Regression in R

In this article, we are going to learn to plot linear regression in R. But, to plot Linear regression, we first need to understand what exactly is linear regression.

What is Linear Regression?

Linear Regression is a supervised learning model, which computes and predicts the output implemented from the linear relationship the model established based on the data it gets fed with. The aim of this model is to find the linear equation that best fits the relationship between the independent variables (features) and the dependent variable (target). This was some basic insight into the concept of Linear Regression. With this, now we can dive into how to plot linear regression in R.

Now we need to understand, why we chose R for this purpose, as we can go with any other language like Python which also has a number of libraries for machine learning and data analysis.

Equation for Linear Regression

The Equation for Linear Regression is very simple which was covered in our primary standards,

y = bx + a

Here,

a: y-Intercept

b: Slope of the regression line (line of best fit)

x: independent variable or features

y: dependent variable or target

Now for a dataset Slope and Intercept can be calculated with the following formulas

Slope = sample covariance/sample variance
Intercept = y_mean – slope* x_mean

Loss Function

Loss function, often referred as Cost Function is basically the difference between the true value (y) and the value predicted by our Machine Learning model (ŷ) which is called error in general terms. This can be calculated using Mean Squared Error method which is a commonly used method in statistics to find out errors.

Calculation of Mean Squared Error

Mean Squared Error or MSE can be calculated by squaring the difference between the predicted and true value for the whole dataset and then calculating its mean by dividing it by the total number of observations

Cost Function = Mean Squared Error =

Why R?

R is a popular programming language that was developed solely for statistical analysis and data visualization. Some of the machine learning models we use are already pre-trained into R (no need for installation of external libraries) including Linear Model (or Regression). Also, this language comes with a huge community and one of the best tools and libraries for Machine Learning, which makes it the first choice for Data Science and Machine Learning enthusiasts.

Now, let's start with plotting linear regression.

Plotting Linear Regression in R

The dataset we are using for this is: placement.csv

Output:

Call:
lm(formula = package ~ cgpa, data = SalaryData)
Residuals:
 Min 1Q Median 3Q Max 
-15.517 -9.794 -4.762 2.376 53.174 
Coefficients:
 Estimate Std. Error t value Pr(>|t|) 
(Intercept) -94.141 47.167 -1.996 0.0613 .
cgpa 17.415 6.704 2.598 0.0182 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 16.56 on 18 degrees of freedom
Multiple R-squared: 0.2726, Adjusted R-squared: 0.2322 
F-statistic: 6.747 on 1 and 18 DF, p-value: 0.01819

👁 placement_stats

• Here, first we are importing the dataset (placement.csv) available in csv (Comma Separated Values) format using read.csv function in R and assigning it to a variable SalaryData.
• Next we are using the pre-built Linear Regression model using , feeding x, y and our dataset to the function as arguments.
• Next, we are creating a png file to store the graph plotted using png function which takes the file name as the argument.
• Finally, we are using the function given by R to plot the graph for the data.

Running this script, will create a file "placement_stats.png" which has the plotted graph for the dataset in the current directory

Changing pch

Output:

Call:
lm(formula = package ~ cgpa, data = SalaryData)
Residuals:
 Min 1Q Median 3Q Max 
-15.517 -9.794 -4.762 2.376 53.174 
Coefficients:
 Estimate Std. Error t value Pr(>|t|) 
(Intercept) -94.141 47.167 -1.996 0.0613 .
cgpa 17.415 6.704 2.598 0.0182 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 16.56 on 18 degrees of freedom
Multiple R-squared: 0.2726, Adjusted R-squared: 0.2322 
F-statistic: 6.747 on 1 and 18 DF, p-value: 0.01819

👁 placement_stats

Changing the pch value from 16 to 17 changed the shape of plotting points from solid circles to triangles, while keeping everything same.

Changing col

Output:

👁 placement_stats

Changing the col from blue to red lead to change in the color of plotting points from blue (initial) to red (new), while keeping everything the same.

Additionally, abline function is used to add the regression line (line of best fit) to the plot.
Removing it will result in a scatterplot.

Removing abline parameter

Output:

👁 placement_stats

When abline parameter is removed from the call of plot function, the line of best fit disappears from the grap, hence making it a scatterplot, while keeping everything the same.

This can also be plotted by using external libraries like ggplot2.

Plotting using ggplot2

First, we need to install the library using following command

install.packages("ggplot2")

• ggplot is a popular data visualization package in R programming.

Then we can use it as usual by importing the library into script.

Output:

👁 placement_stats

It works in similar manner as previous one, both line of best fit and points are plotted on the graph

Without geom_point (plotting points absent)

Output:

👁 placement_stats

When removing geom_point parameter the plotting points are not displayed on the graph while keeping everything same.

Without geom_smooth (line absent)

Output:

👁 placement_stats

When removing geom_smooth parameter the line of best fit is not displayed on the graph while keeping everything same.

Exploring the Relationship Between Hours Studied and Test Scores

Output:

Call:
lm(formula = Test_Score ~ Hours_Studied, data = data)
Residuals:
 1 2 3 4 5 6 7 
-3.03571 -0.64286 1.75000 3.14286 2.53571 -0.07143 -3.67857 
Coefficients:
 Estimate Std. Error t value Pr(>|t|) 
(Intercept) 45.8214 2.9683 15.44 2.07e-05 ***
Hours_Studied 6.6071 0.5512 11.99 7.13e-05 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 2.917 on 5 degrees of freedom
Multiple R-squared: 0.9664, Adjusted R-squared: 0.9596 
F-statistic: 143.7 on 1 and 5 DF, p-value: 7.128e-05

👁 score_stats

A graph has been plotted with the number of Hours studied on the X-axis and Score in the Test on the Y-axis along with a line of best fit.