Linear Regression and Least Squares Worked Examples Cheatsheet and Study Guide

How to start a linear regression and least squares problem without guessing

This worked-examples version of linear regression and least squares is designed to show the order of thought, not just the final result. Worked examples are useful because they expose the order of thought: identify the controlling condition, choose the right model or rule, and only then compute or conclude. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Think of regression as a prediction tool built to make overall vertical error as small as possible for the given data. If you skip that order, even familiar formulas become fragile under slight wording changes. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Study hours and test score model

A scatter plot suggests that higher study time is associated with higher scores and a regression line is fit. The aim here is reading slope and residuals in context. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Use the scatter plot first to justify why a line might be reasonable.
Interpret the slope as the average score change per additional study hour, not as a guaranteed gain for each student.
Then use residuals to explain why individual observations can still sit above or below the prediction.

This example trains the difference between trend and exact prediction. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Residual plot warning sign

A fitted line looks okay at first glance, but the residual plot shows a curved pattern. The aim here is why residuals matter for model checking. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Recognise that a nonrandom residual pattern signals structure the line is missing.
Explain that the relationship may be nonlinear or otherwise mis-specified.
Use the residual evidence to argue for caution rather than forcing a linear interpretation.

This is one of the most useful examples for avoiding blind faith in regression output. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Decision table for recurring linear regression and least squares problems

Problem type	First move	Key check	Typical payoff
Study hours and test score model	Use the scatter plot first to justify why a line might be reasonable.	Interpret the slope as the average score change per additional study hour, not as a guaranteed gain for each student.	This example trains the difference between trend and exact prediction.
Residual plot warning sign	Recognise that a nonrandom residual pattern signals structure the line is missing.	Explain that the relationship may be nonlinear or otherwise mis-specified.	This is one of the most useful examples for avoiding blind faith in regression output.

Patterns the worked examples were meant to teach

Before fitting a line, you need to inspect whether the data show a roughly linear pattern and whether one variable is being used to explain or predict another. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Residuals are the vertical differences between observed and predicted values. The least-squares method chooses the line that makes the sum of squared residuals as small as possible. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Skipping the scatter plot is a common reason a solution feels right while still landing on the wrong conclusion. Inspect the point pattern before trusting any line. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Continue through the linear regression and least squares cluster

Open linear regression and least squares Overview when you want the broad conceptual map before diving back into detail.
Open linear regression and least squares Exam Essentials when you want the highest-yield version of the same topic under time pressure.
This is the page you are already on, so use the note below it as your benchmark for what that variant should deliver.
Open linear regression and least squares Revision Checklist when you want a memory audit instead of another long explanation.
Open linear regression and least squares Common Mistakes when you want to debug the predictable traps that keep appearing in your answers.

Maths pages that reinforce this worked examples

counting principles, permutations, and combinations Worked Examples is the nearest same-variant page if you want a comparable angle on a neighboring maths topic.
confidence intervals for means and proportions Worked Examples is the next same-variant page if you want to keep the revision mode but change the content.
Browse the full maths cheatsheet archive if you want a broader subject sweep after this page.

Linear regression and least squares FAQ for Worked Examples

What makes a regression line ‘best fit’?

The least-squares line is chosen so that the sum of squared residuals is as small as possible for the data. That gives a principled reason for the fitted equation rather than an eyeballed guess. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

What is a residual in plain language?

It is the vertical difference between an observed y-value and the y-value predicted by the regression line at the same x-value. It measures prediction error for that point. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Why is the scatter plot still important if software gives me the line instantly?

Because the plot can reveal outliers, curvature, clustering, or no real relationship at all. A computed line is only as meaningful as the data pattern allows. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Can a strong linear fit prove one variable causes the other?

No. Regression can describe association and support prediction, but causation depends on design, mechanism, and potential confounding variables. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Source trail for linear regression and least squares

OpenStax Introductory Statistics 2e: 12.2 Scatter Plots was used for the regression starts with a scatter plot and a question about relationship framing in this worked examples maths page.
OpenStax Introductory Statistics 2e: 12.3 The Regression Equation was used for the the least-squares line minimizes total squared residual error framing in this worked examples maths page.

Extra consolidation for linear regression and least squares

Think of regression as a prediction tool built to make overall vertical error as small as possible for the given data. That view links the equation, the slope, and the residual plot into one story. A stronger final pass is to connect regression starts with a scatter plot and a question about relationship to the least-squares line minimizes total squared residual error and then force yourself to explain what changes between them instead of memorising each heading in isolation. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Before fitting a line, you need to inspect whether the data show a roughly linear pattern and whether one variable is being used to explain or predict another. Residuals are the vertical differences between observed and predicted values. The least-squares method chooses the line that makes the sum of squared residuals as small as possible. Read those two ideas as one chain and notice how they control the way you would justify the topic in an exam, lab write-up, or data interpretation setting. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

To make that chain usable, walk the process through plot the data first and choose explanatory and response variables. Inspect overall direction, strength, and obvious outliers before calculating a line. Decide which variable is being used to predict the other. The point is not just to know the labels, but to know why this order reduces confusion when the prompt becomes more detailed or wordy. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

A scatter plot suggests that higher study time is associated with higher scores and a regression line is fit. This example trains the difference between trend and exact prediction. Put that beside residual plot warning sign and ask what stays stable across both examples even when the surface details change. That comparison work is usually where durable understanding starts to replace pattern-matching. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

A formula can be computed even when the relationship is nonlinear or driven by an outlier. Inspect the point pattern before trusting any line. Once you can correct that error on purpose, look for interpreting slope without units or context as the next likely point of failure so the topic gets cleaner with each pass instead of just feeling more familiar. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)

Quick recall prompts

Restate regression starts with a scatter plot and a question about relationship in one sentence without leaning on the phrasing already used above. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)
Link that sentence to plot the data first so the topic feels like a sequence of moves instead of a loose list of facts. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)
Rehearse study hours and test score model out loud and ask what evidence or condition you would check first. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)
Scan your next answer for skipping the scatter plot before you decide the response is finished. (OpenStax Introductory Statistics 2e: 12.2 Scatter Plots; OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)
Compare this worked examples page with linear regression and least squares Revision Checklist if you want the same content reframed for a different study task.

This is one of the most useful examples for avoiding blind faith in regression output. If the topic still feels thin after that, move through the sibling and neighboring pages linked above and turn this page into the anchor note that keeps the whole cluster internally connected. (OpenStax Introductory Statistics 2e: 12.3 The Regression Equation)