A residual plot shows how far each observed value is from the value predicted by a model, usually by plotting residuals against the input variable or fitted values.

What it tells you

  • Random scatter around 0 usually means the model fits reasonably well.
  • A clear pattern or curve suggests the model is missing something, such as a non-linear relationship.
  • Changing spread can indicate unequal variance, also called heteroscedasticity.
  • Outliers or unusual points can stand out as large residuals.

In plain language

If a model is good, the residuals should look like random noise. If you can see structure in the plot, that’s a clue the model is not capturing the data well.

If you want, I can also show you how to read a residual plot in 30 seconds with a simple example.