LinearRegression_R

Hello, World! I am the DataByte. This repo hosts my scripts (R) that I walk through in my linear regression series on YouTube.

Check out my video Linear Regression with Excel, Python & R to hear more about these topics:

Mathematics

In my YouTube video, I give a short mathematical explanation of how linear regression trendlines are calculated with the least squares method.

R² value

The r² value (the coefficient of determination) helps you determine if your data is exhibiting linear behavior or not. R² values range between 0 and 1, with 1 indicating a perfect linear trend. Values beteween 0.7 and 1 indicate that your data has a linear trend, the closer to 1, the better. You probably don't want to be using any linear trendlines with r² values less than 0.7 to make predictions.

Code

Linear regression in R is easy using the lm() function.

linearmodel <- lm(y ~ x)

Calling summary() will give you a description of your model.

Linear regression can help you capture the general trend of your data, but if the r² value is less than 0.7, it's not going to be very much good for making predictions.

Sometimes, you may be able to find a subset in your dataset that may still exhibit strongly linear characteristics, like Tesla stock's closing prices between March 14th and April 1st, 2022. It's an interesting observation, but if you told anyone you were trading off a linear model, they'd probably think you were crazy!

Dependencies

None

References

It is easy to find sources online to learn more about linear regression, here are some great references to what I used.

R:

lm: Fitting Linear Models

plot: Generic X-Y Plotting

Mathematical background:

Miller, I., Freund, J. E., & Johnson, R. A. (2005). Miller and Freund's Probability and statistics for engineers. Upper Saddle River, NJ: Prentice Hall.

Sigma summation

More?

If you want to hear more from me, check out my Hello, World! video on YouTube and subscribe! I'm constantly working on new content :)