Normal Probability Plots (Q-Q Plots)
What you'll learn: How to use Q-Q plots to visually check whether your regression residuals follow a normal distribution.
Why Q-Q Plots Matter
One key assumption of linear regression is that residuals are normally distributed. While you could run formal tests, a Normal Probability Plot (also called a Quantile-Quantile plot or Q-Q plot) gives you an immediate visual assessment. It's like holding your data up to a template of perfect normality and seeing where it deviates.
How Q-Q Plots Work
A Q-Q plot compares the quantiles (ordered values) of your residuals against the quantiles you'd expect if the data were perfectly normal.
The process:
- Sort your residuals from smallest to largest
- Calculate what values you'd expect at each position if the data were perfectly normal
- Plot actual residuals (y-axis) versus expected normal quantiles (x-axis)
What to look for:
- Straight diagonal line = residuals are normally distributed ✓
- Curved pattern = skewed distribution (tail problems)
- S-shaped curve = heavy or light tails
- Points far from the line = outliers
Interpretation Example
Think of it like comparing heights in your class to national height standards. If your class is "normally distributed" like the general population, a plot of your class heights versus expected heights should follow a straight line. Deviations tell you where your class differs—maybe more short students (lower tail bends down) or some unusually tall outliers (upper tail bends up).
Key Takeaway: Q-Q plots provide a quick visual check of the normality assumption by plotting observed residual quantiles against theoretical normal quantiles—if the points roughly follow a straight diagonal line, the normality assumption is satisfied.