Statistics for Business: Decision Making and Analysis, 3rd Edition

...

The residuals track the diagonal reference line and remain inside the adjacent bands. These residuals are nearly normal. Fortunately, inferences about β₀ and β₁ work well even if the data are not normally distributed. Confidence intervals for the slope and intercept are reliable even if the errors are not normally distributed. As when finding confidence intervals for μ, this claim relies on the Central Limit Theorem.

If the residuals are not normally distributed, check the sample size condition for the residuals (as in Chapter 14). The sample size should be larger than 10 times the absolute value of the kurtosis K₄ of the residuals.

Modeling ...