A quantile–quantile (Q–Q) plot compares residual quantiles with those of a normal distribution. Points roughly following a 45-degree reference line support normality; systematic departures indicate potential non-normality. section 4.2.4
Use the plot to evaluate regression residuals:
- Inspect the center and tails for departures from the dashed normal-reference line. section 4.2.4
- Treat substantial tail deviations as possible heavy tails or skewness. section 4.2.4
- Follow the subjective visual check with a Jarque–Bera test when a more definitive assessment is needed. section 4.2.4
Python implementation
Run stats.probplot(lm.resid, dist='norm', plot=plt), style the reference line as dashed, title the chart “Q-Q Plot,” and display it with plt.show().
