Popular Courses
Linear Regression
Introduction
Decision Trees
Introduction
Neural Networks
Introduction

Sign in to Curvingo

or

New Account     Reset Password     

Linear Regression
Introduction
How It Works, Briefly
Model Evaluation
Finding the Best Line
Understanding What Our Line Means
Code with Python
Built with Mathigon

Share

Send us Feedback

Please let us know if you have any feedback and suggestions, or if you find any errors and bugs in our content.

Sorry, your message couldn’t be submitted. Please try again!

Thanks for your feedback!

Reset Progress

This will delete your progress and chat data for all chapters in this course, and cannot be undone!

Glossary

Select one of the keywords on the left…

Linear RegressionUnderstanding What Our Line Means

Reading time: ~10 min

The best thing about simple regression is that it's easy to read! The numbers (coefficients) it spits out actually mean something in the real world. Let's look at a few common examples and see how to read them.

Select a tab to see what the numbers mean:

Yes/No Features

Sometimes a feature is just a binary "Yes" (1) or "No" (0). Imagine predicting height based on whether someone plays basketball.

Here, our number simply tells us the average difference between the two groups. If the number is 5, it means basketball players are, on average, 5 inches taller than non-players!

Number Features

What about a continuous number, like predicting ice cream sales based on the temperature outside?

In this case, the number is exactly like the slope from high school math (y=mx+b). If our slope is 10, it means for every 1 degree the temperature goes up, we expect to sell 10 more ice creams!

Multiple Features

When we have many variables at once (like age, weight, and diet), we read the number for "age" as: The change we expect to see for every year you age, assuming all other factors stay exactly the same.

This "all else equal" view is super powerful for isolating what really matters in a messy real-world dataset!

"Team Up" Features

Sometimes variables depend on each other. For example, eating ice cream might only cool you down if the temperature outside is hot!

We handle this by making the variables "team up" (multiplying them together in our math). This means the rule for ice cream changes dynamically depending on what the weather variable is doing.

Of course, this is just scratching the surface, but it's a great start!

The "Unspoken Rules" of Regression

When statisticians talk about regression, they use lots of big words like Homoscedasticity or Normality of Errors. These are just fancy ways to describe the "unspoken rules" our math makes.

Instead of memorizing big academic words, just ask yourself these common-sense questions:

  • Does a straight line even make sense? (If the data looks like a U-shape, drawing a straight line through it is a idea!)
  • Are my errors totally random? (If your line is always guessing too low for tall people and too high for short people, your errors aren't random—you're missing an important !)
  • Did I collect good data? (If you only surveyed teenagers, you shouldn't use your model to predict things for .)

What If We Break the Rules?

Don't panic! In modern machine learning, we mostly care about one thing: Does the model make good predictions on brand new data?

If your predictions are highly accurate on data it has never seen before, then perfectly following the "unspoken rules" isn't always strictly necessary. However, if your predictions are bad, you might need to hunt down which rule you broke and fix your .

Where To Go Next

Regression is a massive topic! There are tricks to stop models from overcomplicating things (Regularization), ways to automate picking the best features, and completely different math formulas for different shapes of data.

For now, give yourself a pat on the back. You've officially tackled the "Hello World" of Machine Learning!

Sina