Identifying Non-Linearity in Residual Plots
What you'll learn: How to spot telltale patterns in residual plots that reveal when a straight-line model doesn't fit your data properly.
The Core Idea
When you plot residuals (the errors) against fitted values, you're looking for randomness. If the residuals scatter randomly around zero with no discernible pattern, your linear model is doing its job. But if you see curves, U-shapes, or any systematic pattern, your straight line is missing something important about the relationship between your variables.
Think of it like drawing a straight line through a curved road on a map. Your "errors" (how far actual road points are from your line) won't be random—they'll follow a pattern. Points at the beginning and end might all be on one side of the line, while points in the middle cluster on the other side.
What to Look For
Curved Patterns
A smooth curve in your residuals suggests the true relationship bends in ways your straight line can't capture. You might see residuals that start positive, dip negative, then go positive again (or vice versa).
U-Shapes or Inverted U-Shapes
These patterns indicate the relationship might be quadratic or polynomial rather than linear. The residuals form a bowl or arch shape instead of scattering randomly.
Why This Matters
If your residuals show these patterns, your predictions will be systematically wrong in predictable ways. This violates the linearity assumption—one of regression's core requirements you've already learned about.
Key Takeaway: Non-random patterns (curves, U-shapes) in residual plots signal that a linear model is inadequate; the relationship requires a more flexible approach or transformation to model correctly.